Virtual power plant scheduling method and device based on large model, and electronic equipment

Abstract

The invention discloses a virtual power plant scheduling method and device based on a large model and electronic equipment, and the method comprises the steps: firstly obtaining a first query instruction of a user through the large model, analyzing a query intention, and obtaining power market transaction data if the query intention is a query power adjustable time period and a price difference thereof; and then, calling a pre-trained time sequence model by the large model, predicting a short-term electricity market electricity price and a relatively long-term electricity market electricity price based on the electricity market transaction data, determining an electricity adjustable time period and an electricity adjustable time period price difference according to the predicted electricity price, and outputting the inquired electricity adjustable time period and the electricity adjustable time period price difference as output results. By utilizing the understanding and scheduling capabilities of a large model, tedious interface interaction is avoided, and a user can ask questions through a natural language to obtain required information. Meanwhile, electricity price prediction is performed by using the pre-trained time sequence model, so that the electric power adjustable time period and the electric power adjustable time period price difference can be identified more effectively, and more accurate resource optimization configuration is realized.

Description

Technical Field

[0001] The present application relates to the field of electric power, and more specifically, to a virtual power plant scheduling method, device, and electronic equipment based on a large model. Background Art

[0002] As a cutting-edge power system management model, virtual power plants have pioneered improvements in the operating efficiency and reliability of power systems by integrating distributed energy, controllable loads, and energy storage resources. The core of this model is to use cutting-edge communication and control technologies to effectively aggregate decentralized energy resources so that they can participate in power market transactions and provide necessary ancillary services. However, despite the potential for the development of virtual power plants, existing virtual power plant scheduling methods face a series of challenges, especially in the context of increasingly diverse user needs. How to formulate the optimal response strategy has become an urgent problem that needs to be solved.

[0003] In addition, traditional virtual power plant dispatch methods rely on dedicated models that operate based on a series of preset rules. At the beginning of the design, these models failed to fully consider the complex interface interactions between the various proprietary models. Therefore, when user needs change, the existing dispatch system often cannot be flexibly adjusted to adapt to the specific needs of users. Users have to access different proprietary models separately, which is not only cumbersome but also inefficient. This limitation not only affects the dispatch efficiency of virtual power plants, but also limits their competitiveness and adaptability in the power market. Summary of the invention

[0004] In view of the above problems, this application is proposed to provide a virtual power plant scheduling method, device, and electronic device based on a large model to improve the scheduling efficiency of the virtual power plant. The specific solution is as follows:

[0005] In a first aspect, a virtual power plant scheduling method based on a large model is provided, comprising:

[0006] Obtaining the user's first query instruction through the big model, and analyzing the query intention of the first query instruction, if the query intention is to query the adjustable power period and its price difference, then obtaining the power market transaction data;

[0007] Calling a pre-trained time series model through the large model to predict a future electricity market price set based on the electricity market transaction data through the time series model, wherein the set includes short-term electricity market prices and relatively long-term electricity market prices;

[0008] Determining the adjustable power period and the price difference of the adjustable power period according to the short-term power market price and the relatively long-term power market price;

[0009] The power adjustable period and the power adjustable period price difference are output as output results.

[0010] Preferably, the method further comprises: obtaining a second query instruction of the user through the large model, and analyzing the query intention of the second query instruction, and if the query intention is to query the device execution plan, obtaining the user's energy storage charge state and the user's energy storage charge and discharge power through the large model;

[0011] The pre-trained reinforcement learning model is called by the large model, and the power market transaction data, the power adjustable period, the power adjustable period price difference, the user energy storage charge state and the user energy storage charge and discharge power are input into the reinforcement learning model to obtain the energy storage decision plan output by the reinforcement learning model that matches the user query. The reinforcement learning model is pre-trained by constructing a reward function with the power market transaction data, the power adjustable period, the power adjustable period price difference and the user energy storage charge state as input states, and the user energy storage charge and discharge power as an execution action with the goal of maximizing benefits;

[0012] Inputting the energy storage decision plan into a preset scheduling optimization solver to obtain the optimal solution for the energy storage decision output by the scheduling optimization solver;

[0013] The large model is called to instruct the large model to obtain a device execution plan that matches the user query based on the optimal solution of the energy storage decision, and the device execution plan is output as an output result. The device execution plan is an adjustment plan and execution instructions for the user's energy storage during the power adjustable period.

[0014] Preferably, the time series model is pre-trained using electricity market transaction data as training samples and a set of future electricity market electricity prices at time nodes corresponding to the electricity market transaction data as training labels.

[0015] Preferably, the process of determining the adjustable power period and the price difference of the adjustable power period according to the short-term power market price and the relatively long-term power market price includes:

[0016] When the difference between the short-term electricity market price and the relatively long-term electricity market price is greater than a preset threshold, the difference is used as the adjustable period price difference, and the time node interval corresponding to the time when the difference between the short-term electricity market price and the relatively long-term electricity market price is greater than the preset threshold is used as the adjustable electricity period.

[0017] Preferably, the short-term electricity market price is the real-time electricity market price, and correspondingly, the relatively long-term electricity market price is the day-ahead electricity market price or the medium- to long-term electricity market price;

[0018] or,

[0019] The short-term electricity market price is the day-ahead electricity price in the electricity market, and correspondingly, the relatively long-term electricity market price is the medium- and long-term electricity price in the electricity market.

[0020] Preferably, the execution action of the training process of the reinforcement learning model further includes constraints in addition to the charging and discharging power of the user's energy storage, and the constraints include:

[0021] The capacity of the user energy storage cannot be lower than the minimum capacity of the user energy storage or exceed the maximum capacity of the user energy storage, the power market load demand cannot exceed the charging and discharging capacity of the user energy storage, the user energy storage cannot be charged and discharged at the same time, and the net change of the capacity of the user energy storage plus charging and discharging cannot exceed the maximum capacity of the user energy storage.

[0022] Preferably, the electricity market transaction data includes: electricity market load demand, electricity market controllable load information, and / or power grid operation status information.

[0023] Preferably, the reinforcement learning model is a Q-learning model, a proximal policy optimization PPO model or a deep Q network DQN model.

[0024] In a second aspect, a virtual power plant dispatching device based on a large model is provided, comprising:

[0025] A first instruction response unit is used to obtain a first query instruction of a user through a large model, and analyze the query intention of the first query instruction. If the query intention is to query the adjustable power period and its price difference, then obtain power market transaction data;

[0026] A time series model calling unit, configured to call a pre-trained time series model through the large model, so as to predict a future power market price set based on the power market transaction data through the time series model, wherein the set includes a short-term power market price and a relatively long-term power market price;

[0027] A data analysis unit, configured to determine an adjustable power period and a price difference of the adjustable power period according to the short-term power market price and the relatively long-term power market price;

[0028] The first intention result output unit is used to output the adjustable power period and the price difference of the adjustable power period as output results.

[0029] In a third aspect, an electronic device is provided, comprising: a memory and a processor;

[0030] The memory is used to store programs;

[0031] The processor is used to execute the program to implement the various steps of the virtual power plant scheduling method described in any one of the first aspects of the present application.

[0032] By means of the above technical solution, the present application uses a large model to provide an interactive interface with the user. The user can interact with the large model by asking questions. The large model, as a core control unit, can call the model to execute the corresponding sub-functions to obtain the corresponding answers to the user's questions, avoiding the problem of complicated interface interaction. The large model can obtain the user's first query instruction and analyze the query intention of the first query instruction. If the query intention is the adjustable power period and its price difference, the large model calls the pre-trained timing model to analyze the power market transaction data, which can more accurately obtain the short-term power market electricity price and the relatively long-term power market electricity price. According to the above electricity prices, the adjustable power period and the adjustable power period price difference are obtained and output, realizing flexible resource optimization configuration, which can effectively solve the problem that traditional dedicated models are difficult to adapt to different user needs and market changes, and improve the response speed and flexibility of virtual power plants. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present application. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0034] Figure 1 A schematic diagram of an implementation system architecture of a large model-based virtual power plant scheduling method provided in an embodiment of the present application;

[0035] Figure 2 A schematic diagram of a terminal structure provided in an embodiment of the present application;

[0036] Figure 3 A schematic diagram of a server structure provided in an embodiment of the present application;

[0037] Figure 4 A schematic diagram of a process of a virtual power plant scheduling method based on a large model provided in an embodiment of the present application;

[0038] Figure 5 Another schematic diagram of a flow chart of a virtual power plant scheduling method based on a large model provided in an embodiment of the present application;

[0039] Figure 6 A scenario diagram illustrating a virtual power plant scheduling method based on a large model;

[0040] Figure 7 A schematic diagram of the structure of a virtual power plant dispatching device based on a large model provided in an embodiment of the present application;

[0041] Figure 8 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0043] First of all, the method provided in this application involves concepts related to timing models and reinforcement learning models. For ease of understanding, the relevant concepts of the timing models and reinforcement learning models involved are introduced below.

[0044] Time series model: A time series model, also known as a time series model, is a statistical model that predicts future results based on historical data. The core idea is to use past information to infer future trends or patterns. Time series models are particularly suitable for scenarios where there is time dependency between data points. For example, autoregressive model (AR), moving average model (MA), autoregressive moving average model (ARMA), etc. In a time series model, the training and test splits of the data are performed in chronological order to ensure that the model can capture the time dependency of time series data. Time series models are widely used in financial market analysis, weather forecasting, sales forecasting and other fields to help decision makers make more accurate future predictions based on historical data.

[0045] Reinforcement Learning Model: Reinforcement Learning (RL) is a paradigm of machine learning, which is mainly used to describe and solve the problem of how an agent learns strategies to maximize rewards or achieve specific goals during the interaction with the environment. The core of the reinforcement learning model lies in the continuous interaction between the agent and the environment. The agent obtains rewards by taking actions, and then optimizes its own action strategy in the hope of maximizing its long-term benefits. Reinforcement learning can usually be represented by a Markov decision process (MDP), whose main elements include state (S), action (A), reward (R), state transition probability (T) and discount factor (γ). The agent's strategy π represents a mapping from state space to action space, and the goal is to optimize the strategy to obtain the maximum reward value. Reinforcement learning models are widely used in information theory, game theory, automatic control and other fields. They are used to explain equilibrium states under bounded rationality conditions, design recommendation systems and robot interaction systems, etc. In recent years, deep reinforcement learning (DRL) combines the advantages of deep learning and reinforcement learning, and is mainly used to solve perception-decision problems, becoming an important research branch in the field of artificial intelligence.

[0046] This application provides a virtual power plant scheduling method based on a large model, which can be applied to Figure 1 The system architecture shown in FIG. 1 may include a terminal 100 and a server 200. The server 200 may include one or more servers ( Figure 1 A server is included as an example for explanation).

[0047] The terminal 100 or the server 200 can be used alone to execute the virtual power plant scheduling method based on a large model provided in the embodiment of the present application. In addition, the terminal 100 and the server 200 can also be used together to execute the virtual power plant scheduling method based on a large model provided in the embodiment of the present application.

[0048] Next describe Figure 1 The product form of the mid-terminal 100;

[0049] The terminal 100 in the embodiment of the present application can be a mobile phone, a tablet computer, a conference terminal, an augmented reality (AR) / virtual reality (VR) device, a laptop computer, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (PDA), etc., and the embodiment of the present application does not impose any limitation on this.

[0050] Figure 2 An optional hardware structure diagram of the terminal 100 is shown.

[0051] refer to Figure 2 As shown, the terminal 100 may include a radio frequency unit 110, a memory 120, an input unit 130, a display unit 140, a camera 150 (optional), an audio circuit 160, a speaker 161, a microphone 162, an earphone jack 163 (optional), a processor 170, an external interface 180, a power supply 190 and other components. Those skilled in the art will appreciate that Figure 2 These are merely examples of terminals or multi-function devices and do not constitute limitations on the terminals or multi-function devices, which may include more or fewer components than those shown in the figures, or combinations of certain components, or different components.

[0052] The input unit 130 can be used to receive input digital or character information, and generate key signal input related to the user settings and function control of the portable multifunctional device. Specifically, the input unit 130 may include a touch screen 131 and / or other input devices 132. The touch screen 131 can collect the user's touch operations on or near it (such as the user's operation on or near the touch screen using any suitable object such as fingers, joints, stylus, etc.), and drive the corresponding connection device according to a pre-set program. The touch screen can detect the user's touch action on the touch screen, convert the touch action into a touch signal and send it to the processor 170, and can receive and execute the command sent by the processor 170; the touch signal at least includes the touch point coordinate information. The touch screen 131 can provide an input interface and an output interface between the terminal 100 and the user. In addition, the touch screen can be implemented using multiple types such as resistive, capacitive, infrared and surface acoustic wave. In addition to the touch screen 131, the input unit 130 can also include other input devices. Specifically, other input devices 132 may include, but are not limited to, one or more of a physical keyboard, function keys (such as a volume control key, a switch key, etc.), a trackball, a mouse, a joystick, and the like.

[0053] Among them, other input devices 132 can receive input data and so on.

[0054] The display unit 140 may be used to display information input by the user or provided to the user, various menus of the terminal 100, interactive interfaces, file display, and / or playback of any multimedia file. In an embodiment of the present application, the display unit 140 may be used to display various interactive interfaces, processing results, etc. in a virtual power plant scheduling method based on a large model.

[0055] The memory 120 can be used to store instructions and data. The memory 120 can mainly include an instruction storage area and a data storage area. The data storage area can store various data, such as multimedia files, texts, etc.; the instruction storage area can store software units such as operating systems, applications, instructions required for at least one function, or their subsets and extensions. It can also include a non-volatile random access memory; provide the processor 170 with hardware, software and data resources including management of computing and processing equipment, and support control software and applications. It is also used for the storage of multimedia files, and the storage of running programs and applications.

[0056] The processor 170 is the control center of the terminal 100. It uses various interfaces and lines to connect various parts of the entire terminal 100. By running or executing instructions stored in the memory 120 and calling data stored in the memory 120, it executes various functions of the terminal 100 and processes data, thereby controlling the terminal device as a whole. Optionally, the processor 170 may include one or more processing units; preferably, the processor 170 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface and application program, and the modem processor mainly processes wireless communication. It is understandable that the above-mentioned modem processor may not be integrated into the processor 170. In some embodiments, the processor and the memory may be implemented on a single chip, and in some embodiments, they may also be implemented separately on separate chips. The processor 170 may also be used to generate corresponding operation control signals, send them to corresponding components of the computing and processing device, read and process data in the software, especially read and process data and programs in the memory 120, so that each functional module therein performs corresponding functions, thereby controlling the corresponding components to act according to the requirements of the instructions.

[0057] Among them, the memory 120 can be used to store software codes related to the virtual power plant scheduling method based on a large model, the processor 170 can execute the steps of the virtual power plant scheduling method based on a large model, and can also schedule other units (such as the above-mentioned input unit 130 and the display unit 140) to achieve corresponding functions.

[0058] The radio frequency unit 110 (optional) can be used for receiving and sending information or receiving and sending signals during a call, for example, after receiving the downlink information of the base station, it is sent to the processor 170 for processing; in addition, the designed uplink data is sent to the base station. Generally, the RF circuit includes but is not limited to an antenna, at least one amplifier, a transceiver, a coupler, a low noise amplifier (Low Noise Amplifier, LNA), a duplexer, etc. In addition, the radio frequency unit 110 can also communicate with network devices and other devices through wireless communication. The wireless communication can use any communication standard or protocol, including but not limited to Global System of Mobile communication (Global System of Mobile communication, GSM), General Packet Radio Service (General Packet Radio Service, GPRS), Code Division Multiple Access (Code Division Multiple Access, CDMA), Wideband Code Division Multiple Access (Wideband Code Division Multiple Access, WCDMA), Long Term Evolution (Long Term Evolution, LTE), email, Short Messaging Service (SMS), etc.

[0059] In this embodiment of the present application, the RF unit 110 can send data to the server 200 and receive the processing result sent by the server 200. Exemplarily, the RF unit 110 sends the received user query to the server 200, and the server 200 obtains the output result corresponding to the question information and returns the output result to the terminal 100.

[0060] It should be understood that the radio frequency unit 110 is optional and can be replaced by other communication interfaces, such as a network port.

[0061] The terminal 100 also includes a power supply 190 (such as a battery) for supplying power to various components. Preferably, the power supply can be logically connected to the processor 170 through a power management system, so that the power management system can manage functions such as charging, discharging, and power consumption.

[0062] The terminal 100 further includes an external interface 180 , which may be a standard Micro USB interface or a multi-pin connector, and may be used to connect the terminal 100 to communicate with other devices, or to connect a charger to charge the terminal 100 .

[0063] Although not shown, the terminal 100 may also include a flashlight, a wireless fidelity (WiFi) module, a Bluetooth module, sensors with different functions, etc., which are not described in detail here. Some or all of the methods described below may be applied in the following embodiments. Figure 2 In the terminal 100 shown.

[0064] Next describe Figure 1 The product form of the server 200;

[0065] Figure 3 A schematic diagram of the structure of a server 200 is provided, such as Figure 3 As shown, the server 200 includes a bus 201, a processor 202, a communication interface 203, and a memory 204. The processor 202, the memory 204, and the communication interface 203 communicate with each other via the bus 201.

[0066] The bus 201 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0067] The processor 202 may be any one or more of a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor (MP), or a digital signal processor (DSP).

[0068] The memory 204 may include a volatile memory, such as a random access memory (RAM). The memory 204 may also include a non-volatile memory, such as a read-only memory (ROM), a flash memory, a hard disk drive (HDD), or a solid state drive (SSD).

[0069] Among them, the memory 204 can be used to store software codes related to the virtual power plant scheduling method, the processor 202 can execute the steps of the virtual power plant scheduling method, and can also schedule other units to achieve corresponding functions.

[0070] It should be understood that the above-mentioned terminal 100 and server 200 can be centralized or distributed devices, and the processors in the above-mentioned terminal 100 and server 200 (such as processor 170 and processor 202) can be hardware circuits (such as application specific integrated circuit (ASIC), field-programmable gate array (FPGA), general-purpose processor, digital signal processor (DSP), microprocessor or microcontroller, etc.), or a combination of these hardware circuits. For example, the processor can be a hardware system with an instruction execution function, such as a CPU, DSP, etc., or a hardware system without an instruction execution function, such as an ASIC, FPGA, etc., or a combination of the above-mentioned hardware systems without an instruction execution function and hardware systems with an instruction execution function.

[0071] The embodiment of the present application provides a virtual power plant scheduling method based on a large model, and the method is applied to a computer device as an example. The computer device can be specifically Figure 1 The terminal 100 or the system consisting of the terminal 100 and the server 200. Figure 4 The virtual power plant scheduling method based on the large model specifically includes the following steps:

[0072] Step S100: obtaining a first query instruction of the user through a large model, and analyzing the query intention of the first query instruction. If the query intention is to query the adjustable power period and its price difference, obtaining power market transaction data.

[0073] Specifically, the user interacts with the big model through natural language and issues the first query instruction. The big model will first perform semantic understanding and analysis on this instruction to accurately identify the user's query intention. If the big model determines that the user's intention is to query the adjustable power time period and its price difference information, for example, the user asks "which time periods can be used for power adjustment and the corresponding price differences", the big model will start the subsequent process and obtain power market transaction data from relevant data sources (such as power trading platforms, energy management systems, etc.). Power market transaction data may include power market load demand, power market controllable load information, and / or, power grid operation status information, etc.

[0074] Step S110, calling a pre-trained time series model through the large model to predict a future electricity market price set based on the electricity market transaction data through the time series model, wherein the set includes short-term electricity market prices and relatively long-term electricity market prices.

[0075] Specifically, after obtaining relevant data, the big model will call a pre-trained time series model. This time series model can be a long short-term memory network (LSTM), a recurrent neural network (RNN) or other models suitable for processing time series data. The big model inputs the extracted power market transaction data into the time series model to obtain the predicted short-term power market electricity price and the relatively long-term power market electricity price. The time series model is pre-trained with power market transaction data as training samples and the future power market electricity price set corresponding to the time node of the power market transaction data as training labels.

[0076] For example, the short-term electricity market price can be the average electricity price of the next 24 hours, while the relatively long-term electricity market price can be the average electricity price of the next week. Alternatively, the short-term electricity market price can be the real-time electricity price, while the relatively long-term electricity market price can be the average electricity price of the next 24 hours. "Electricity market price set" refers to a set of electricity price forecast results of different time granularities, and its short-term electricity market price and long-term electricity market price are relative concepts, not limited to the above examples.

[0077] Step S120: determining the power adjustable period and the price difference of the power adjustable period according to the short-term power market price and the relatively long-term power market price.

[0078] Specifically, by analyzing the short-term electricity market electricity price and the relatively long-term electricity market electricity price, the adjustable electricity period and the price difference of the adjustable electricity period are determined. The adjustable electricity period refers to the period when the electricity price is relatively low and the user can flexibly adjust the electricity consumption. For example, under the peak-valley electricity price mechanism, the electricity price during the peak period is usually lower and can be considered as the adjustable electricity period. The price difference of the adjustable electricity period can refer to the difference in electricity price between the adjustable electricity period and other periods. Calculating the price difference of the adjustable electricity period can help users quantify the potential benefits of adjusting electricity consumption. For example, comparing the difference in electricity prices between the peak period and the valley period can help users decide whether to transfer part of the electricity load to the valley period.

[0079] Step S130: output the adjustable power period and the price difference of the adjustable power period as output results.

[0080] Specifically, the large model returns the adjustable electricity period and the price difference between the adjustable electricity periods determined in step 120 to the user as output results. The output results can be presented in a variety of ways, such as charts, reports or directly displayed on the user interface, so that the user can clearly see the adjustable electricity period and the corresponding adjustable electricity period price difference. This step can provide customized information according to the actual needs of the user, thereby enhancing the user experience.

[0081] In one possible implementation, the model called by the large model is not limited to the time series model. According to the user's query intention, the corresponding model or program function can be called for analysis, such as a load model, a power generation model, etc. That is, the method provided in this application focuses on the core control function of the large model. The role of the large model is to call the model or program function to execute its corresponding sub-functions to obtain an answer that meets the user's query, thereby avoiding complicated interface interactions.

[0082] The virtual power plant scheduling method provided in this application realizes efficient management and optimized scheduling of power resources by integrating advanced data analysis and prediction technologies, improves the operating efficiency and reliability of the power system, and also provides users with more accurate and personalized services.

[0083] The above embodiment introduces that when the user's query intention is to query the adjustable power period and its price difference, the output result is the adjustable power period and the price difference of the adjustable power period. On this basis, the user may interact with the large model to continue to query the device execution plan. This embodiment introduces that when the user's query intention is to query the device execution plan, the specific implementation process of the virtual power plant scheduling method can also include the following steps based on the above embodiment:

[0084] Step S200: obtain the user's second query instruction through the big model, and analyze the query intention of the second query instruction. If the query intention is to query the device execution plan, obtain the user's energy storage charge state and the charging and discharging power of the user's energy storage through the big model.

[0085] Specifically, after the user completes the first round of interaction with the big model, he or she can continue to issue a second query instruction. The big model performs semantic understanding and analysis on the second query instruction to accurately identify the user's query intent. If the big model determines that the user's intention is to query the device execution plan, the big model will further obtain the state of charge (SOC) of the user's energy storage related to the user's energy storage system, as well as the charge and discharge power of the user's energy storage. This information is crucial for formulating a reasonable energy storage charge and discharge plan. The big model can access the user's energy storage data, or obtain this information through user input.

[0086] Step S210: calling a pre-trained reinforcement learning model through the large model, inputting the power market transaction data, power adjustable time period, power adjustable time period price difference, user energy storage charge state and user energy storage charge and discharge power into the reinforcement learning model, and obtaining the energy storage decision plan output by the reinforcement learning model that matches the user query.

[0087] Specifically, the large model calls a pre-trained reinforcement learning model. This reinforcement learning model takes the power market transaction data, power adjustable time period, power adjustable time period price difference and user energy storage charge state as input states in advance, takes the charge and discharge power of the user energy storage as the execution action, and constructs a reward function for training with the goal of maximizing benefits, for example, by charging in low-price periods and discharging in high-price periods to reduce electricity costs or gain benefits. The large model inputs the user energy storage charge state and charge and discharge power obtained in step S200, as well as the power market transaction data, power adjustable time period and price difference information obtained previously into the reinforcement learning model. The reinforcement learning model will output an energy storage decision plan based on these input information. This plan may include a series of charging and discharging actions in different time periods, such as "charging at maximum power from 0 to 6 o'clock, and discharging at maximum power from 18 to 22 o'clock". Among them, the reinforcement learning model can be a Q-Learning Model (Q-learning), a proximal policy optimization model (PPO) or a deep Q-network model (DQN), etc., but is not limited to the above examples.

[0088] Step S220: input the energy storage decision plan into a preset scheduling optimization solver to obtain the optimal solution for the energy storage decision output by the scheduling optimization solver.

[0089] Specifically, the energy storage decision plan output by the reinforcement learning model may only be a preliminary plan. In order to further optimize this plan, it is input into a preset scheduling optimization solver. The scheduling optimization solver can consider more constraints, such as battery life, charging and discharging times, etc., and fine-tune the output of the reinforcement learning model to finally obtain an optimal solution for energy storage decision-making. This optimal solution is an energy storage charging and discharging plan that can maximize user benefits while satisfying all constraints.

[0090] Step S230, calling the large model to instruct the large model to obtain a device execution plan that matches the user query based on the optimal solution of the energy storage decision, and output the device execution plan as an output result, wherein the device execution plan is an adjustment plan and execution instructions for the user's energy storage during the power adjustable period.

[0091] Specifically, the big model converts the optimal solution for energy storage decisions output by the scheduling optimization solver into an equipment execution plan for the user's energy storage. This plan can clearly instruct the user in which time periods to charge or discharge, as well as the specific charging and discharging power. For example, the plan may contain instructions such as "charge at 5kW from 0 to 6 o'clock, and discharge at 10kW from 18 to 22 o'clock." The big model will return this equipment execution plan to the user as the final output result, and the user can operate their energy storage system according to this plan.

[0092] By integrating advanced data analysis, reinforcement learning and optimization solution technologies, intelligent management and optimized scheduling of users' energy storage equipment are achieved, the utilization efficiency and economic benefits of energy storage resources are improved, and at the same time, more accurate and efficient equipment execution plans are provided to users.

[0093] The following examples refer to Figure 6 , a scenario diagram illustrating a virtual power plant scheduling method based on a large model.

[0094] like Figure 6 As shown in the figure, a scenario of the virtual power plant scheduling method is described. First, the user initiates a query request by interacting with the big model. The user can express the demand in natural language. After receiving the user's instruction, the big model will analyze the user's query intention.

[0095] If the user wants to query the adjustable power period and its price difference, the big model will obtain power market transaction data from the data layer, such as power market load demand, controllable load information, and power grid operation status. Then, the big model will call the pre-trained time series model to predict future power market prices, including short-term power market prices and relatively long-term power market prices. After obtaining the predicted prices, the big model will analyze short-term power market prices and relatively long-term power market prices, determine the adjustable power period and the price difference of the adjustable power period, and return this information to the user as a result.

[0096] If the user wants to further query the device execution plan, the big model will continue to interact with the user to obtain information such as the state of charge and charge and discharge power of the user's energy storage system. Next, the big model will call the pre-trained reinforcement learning model. The reinforcement learning model will take the power market transaction data, the adjustable power period, the price difference of the adjustable power period, the user's energy storage state of charge and charge and discharge power as input, and output a storage decision plan. In order to further optimize this plan, the big model will input the output of the reinforcement learning model into a scheduling optimization solver. The scheduling optimization solver will consider more constraints, fine-tune the energy storage decision plan, and finally obtain an optimal solution for the energy storage decision. Finally, the big model will convert this optimal solution into a user-understandable device execution plan and return the device execution plan to the user.

[0097] The above-mentioned virtual power plant scheduling method simplifies user operations through natural language interaction, obtains and analyzes electricity market data in real time, and uses pre-trained time series models, reinforcement learning models and scheduling optimization solvers to obtain output results that meet user queries, ensuring scheduling efficiency and improving user experience and the response speed of the virtual power plant.

[0098] In this embodiment, the process of determining the adjustable power period and the price difference of the adjustable power period according to the short-term power market price and the relatively long-term power market price in step S120 in the above embodiment is described in detail, which may include the following steps:

[0099] When the difference between the short-term electricity market price and the relatively long-term electricity market price is greater than a preset threshold, the difference is used as the adjustable period price difference, and the time node interval corresponding to the time when the difference between the short-term electricity market price and the relatively long-term electricity market price is greater than the preset threshold is used as the adjustable electricity period.

[0100] Specifically, for example, the short-term electricity market price may be the hourly electricity price in the next 24 hours, and the relatively long-term electricity market price may be the daily average electricity price in the next week. These two electricity prices can be compared for the same time period. For example, the hourly electricity price in the next 24 hours is compared with the average electricity price in the next 24 hours, or the daily average electricity price in the next week is compared with the overall average electricity price in the next week. Then, the short-term electricity market price and the relatively long-term electricity market price in the same time period are compared point by point to calculate their difference.

[0101] Next, a preset threshold is introduced. This threshold can be adjusted according to the user's risk preference and market fluctuations. If the difference between the short-term electricity market price and the relatively long-term electricity market price at a certain time point is greater than the preset threshold, it is considered that there is an obvious price difference at that time point and it can be regulated. If the difference is greater than the preset threshold, the difference is recorded as the adjustable period price difference. At the same time, the time node interval corresponding to the difference between the short-term electricity market price and the relatively long-term electricity market price is greater than the preset threshold is used as the electricity adjustable period. For example, if between 2 pm and 4 pm, the short-term electricity market price is significantly higher than the relatively long-term electricity market price, and the difference is greater than the preset threshold, this period is determined as the electricity adjustable period, and the corresponding difference is recorded as the adjustable period price difference.

[0102] By introducing the threshold, the adjustable power period with significant price difference can be more effectively identified, avoiding overreaction to small price difference, thus improving the efficiency and economic benefits of energy storage regulation. At the same time, it is also more flexible and can adjust the threshold according to different market conditions and user needs to achieve the best regulation effect.

[0103] The following embodiments further supplement the training process of the reinforcement learning model in the above embodiments, and may specifically include the following steps:

[0104] The execution actions of the training process of the reinforcement learning model include not only the charging and discharging power of the user's energy storage, but also constraints, which include:

[0105] The capacity of the user energy storage cannot be lower than the minimum capacity of the user energy storage or exceed the maximum capacity of the user energy storage, the power market load demand cannot exceed the charging and discharging capacity of the user energy storage, the user energy storage cannot be charged and discharged at the same time, and the net change of the capacity of the user energy storage plus charging and discharging cannot exceed the maximum capacity of the user energy storage.

[0106] Specifically, the execution action of the reinforcement learning model contains the charging and discharging power of the user's energy storage and an indication of satisfying the constraints. The execution action can be defined as a tuple: (power, action type), where: power represents the charging and discharging power, which can be positive (discharging), negative (charging) or zero (neither charging nor discharging). The action type represents the type of action, such as "charging", "discharging" or "maintaining". The introduction of action types can express the model's decisions more clearly and facilitate subsequent constraint checks. Before executing an action, the following constraints can be checked: Capacity limit: The user's energy storage capacity cannot be lower than the minimum capacity or exceed the maximum capacity. Let the current energy storage capacity be SOC_current, the minimum capacity be SOC_min, and the maximum capacity be SOC_max, then it must satisfy: SOC_min<=SOC_current<=SOC_max. Charging and discharging capacity limit: The load demand in the power market cannot exceed the charging and discharging capacity of the user's energy storage. Assume that the load demand of the electricity market is Load, the maximum discharge power of the energy storage is P_discharge_max, and the maximum charging power of the energy storage is P_charge_max, then it must satisfy: -P_charge_max<=Load<=P_discharge_max. Simultaneous charging and discharging restriction: The user's storage cannot be charged and discharged at the same time. This means that if the action type is "charging", the power must be negative or zero; if the action type is "discharging", the power must be positive or zero. Capacity change restriction: The capacity of the energy storage plus the net change of charging and discharging cannot exceed the maximum capacity. Assume that the power of the action is P and the time interval is Δt, then the change in the energy storage capacity is ΔSOC=P×Δt. The new energy storage capacity SOC_new=SOC_current+ΔSOC needs to satisfy: SOC_new<=SOC_max. Similarly, the new energy storage capacity cannot be lower than the minimum capacity: SOC_new>=SOC_min. When all constraints are met, the selected action is executed. If the selected action does not satisfy the constraints, appropriate measures need to be taken. For example, the model can choose an alternative action that satisfies the constraints, choose to skip the current step and not perform any action, or during the reinforcement learning training process, the behavior that violates the constraints can be punished to guide the model to learn a strategy that meets the constraints.

[0107] By introducing these constraints, we can ensure that the strategies learned by the reinforcement learning model are safe and feasible, and can better meet the needs of practical applications. At the same time, more sophisticated action definitions and state space designs can improve the learning efficiency and control accuracy of the model.

[0108] The following describes a large-model-based virtual power plant scheduling device provided in an embodiment of the present application. The large-model-based virtual power plant scheduling device described below and the large-model-based virtual power plant scheduling method described above can refer to each other.

[0109] See also Figure 7 , Figure 7 A schematic diagram of the structure of a virtual power plant scheduling device based on a large model disclosed in an embodiment of the present application.

[0110] like Figure 7 As shown, the device may include:

[0111] The first instruction response unit 11 is used to obtain the user's first query instruction through the large model, and analyze the query intention of the first query instruction. If the query intention is to query the adjustable power period and its price difference, then obtain the power market transaction data;

[0112] A time series model calling unit 12 is used to call a pre-trained time series model through the large model, so as to predict a future power market price set based on the power market transaction data through the time series model, wherein the set includes a short-term power market price and a relatively long-term power market price;

[0113] A data analysis unit 13, configured to determine an adjustable power period and a price difference of the adjustable power period according to the short-term power market price and the relatively long-term power market price;

[0114] The first intention result output unit 14 is used to output the power adjustable period and the power adjustable period price difference as output results.

[0115] Optionally, the device of the present application may further include:

[0116] The second instruction response unit 21 is used to obtain the user's second query instruction through the large model, and analyze the query intention of the second query instruction. If the query intention is to query the device execution plan, the user's energy storage charge state and the user's energy storage charge and discharge power are obtained through the large model;

[0117] A reinforcement learning calling unit 22 is used to call a pre-trained reinforcement learning model through the large model, input the power market transaction data, the power adjustable period, the power adjustable period price difference, the user energy storage charge state and the user energy storage charge and discharge power into the reinforcement learning model, and obtain the energy storage decision plan output by the reinforcement learning model that matches the user query. The reinforcement learning model pre-uses the power market transaction data, the power adjustable period, the power adjustable period price difference and the user energy storage charge state as input states, and uses the user energy storage charge and discharge power as an execution action, and constructs a reward function with the goal of maximizing benefits for training;

[0118] A solver calling unit 23 is used to input the energy storage decision scheme into a preset scheduling optimization solver to obtain the optimal solution of the energy storage decision output by the scheduling optimization solver;

[0119] The second intention result output unit 24 is used to call the large model to instruct the large model to obtain a device execution plan that matches the user query based on the optimal solution of the energy storage decision, and output the device execution plan as an output result. The device execution plan is an adjustment plan and execution instructions for the user's energy storage during the power adjustable period.

[0120] Optionally, the time series model is pre-trained using electricity market transaction data as training samples and a set of future electricity market electricity prices at time nodes corresponding to the electricity market transaction data as training labels.

[0121] Optionally, the process of the data analysis unit 13 determining the power adjustable period and the power adjustable period price difference according to the short-term power market price and the relatively long-term power market price includes:

[0122] When the difference between the short-term electricity market price and the relatively long-term electricity market price is greater than a preset threshold, the difference is used as the adjustable period price difference, and the time node interval corresponding to the time when the difference between the short-term electricity market price and the relatively long-term electricity market price is greater than the preset threshold is used as the adjustable electricity period.

[0123] Optionally, the short-term electricity market price is a real-time electricity market price, and correspondingly, the relatively long-term electricity market price is a day-ahead electricity market price or a medium- to long-term electricity market price;

[0124] or,

[0125] The short-term electricity market price is the day-ahead electricity price in the electricity market, and correspondingly, the relatively long-term electricity market price is the medium- and long-term electricity price in the electricity market.

[0126] Optionally, the execution action of the training process of the reinforcement learning model further includes constraints in addition to the charging and discharging power of the user energy storage, and the constraints include:

[0127] The capacity of the user energy storage cannot be lower than the minimum capacity of the user energy storage or exceed the maximum capacity of the user energy storage, the power market load demand cannot exceed the charging and discharging capacity of the user energy storage, the user energy storage cannot be charged and discharged at the same time, and the net change of the capacity of the user energy storage plus charging and discharging cannot exceed the maximum capacity of the user energy storage.

[0128] Optionally, the electricity market transaction data includes: electricity market load demand, electricity market controllable load information, and / or power grid operation status information.

[0129] Optionally, the reinforcement learning model is a Q-learning model, a proximal policy optimization PPO model or a deep Q network DQN model.

[0130] The present application also provides an electronic device in an embodiment. Figure 8 As shown, it shows a schematic diagram of the structure of an electronic device suitable for implementing the embodiment of the present application. The electronic device in the embodiment of the present application may include but is not limited to fixed terminals such as mobile phones, tablet computers, teaching large screens, wearable devices, etc. Figure 8 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0131] like Figure 8 As shown, the electronic device may include a processing device (such as a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 to a random access memory (RAM) 603 to implement the virtual power plant scheduling method of the aforementioned embodiment of the present application. When the electronic device is powered on, various programs and data required for the operation of the electronic device are also stored in the RAM 603. The processing device 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0132] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a memory card, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device to communicate with other devices wirelessly or by wire to exchange data. Although Figure 8 An electronic device having various devices is shown, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.

[0133] It should also be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed over multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the drawings of the device embodiments provided by the present application, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines.

[0134] Through the description of the above implementation mode, the technicians in the field can clearly understand that the present application can be implemented by means of software plus necessary general hardware, and of course, it can also be implemented by special hardware including special integrated circuits, special CPUs, special memories, special components, etc. In general, all functions completed by computer programs can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be various, such as analog circuits, digital circuits or special circuits. However, for the present application, software program implementation is a better implementation mode in more cases. Based on such an understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a readable storage medium, such as a computer floppy disk, a U disk, a mobile hard disk, a ROM, a RAM, a disk or an optical disk, etc., including a number of instructions to enable a computer device (which can be a personal computer, a training device, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0135] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product.

[0136] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, a computer, a training device, or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, training device, or data center. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device, a data center, etc. that includes one or more available media integrations. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)), etc.

[0137] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can refer to each other.

Claims

1. A virtual power plant scheduling method based on a large model, characterized in that: include: Obtaining the user's first query instruction through the big model, and analyzing the query intention of the first query instruction, if the query intention is to query the adjustable power period and its price difference, then obtaining the power market transaction data; Calling a pre-trained time series model through the large model to predict a future electricity market price set based on the electricity market transaction data through the time series model, wherein the set includes short-term electricity market prices and relatively long-term electricity market prices; Determining the adjustable power period and the price difference of the adjustable power period according to the short-term power market price and the relatively long-term power market price; The power adjustable period and the power adjustable period price difference are output as output results.

2. The method according to claim 1, characterized in that Also includes: The user's second query instruction is obtained through the large model, and the query intention of the second query instruction is analyzed. If the query intention is to query the device execution plan, the user's energy storage charge state and the user's energy storage charge and discharge power are obtained through the large model; The pre-trained reinforcement learning model is called by the large model, and the power market transaction data, the power adjustable period, the power adjustable period price difference, the user energy storage charge state and the user energy storage charge and discharge power are input into the reinforcement learning model to obtain the energy storage decision plan output by the reinforcement learning model that matches the user query. The reinforcement learning model is pre-trained by constructing a reward function with the power market transaction data, the power adjustable period, the power adjustable period price difference and the user energy storage charge state as input states, and the user energy storage charge and discharge power as an execution action with the goal of maximizing benefits; Inputting the energy storage decision plan into a preset scheduling optimization solver to obtain the optimal solution for the energy storage decision output by the scheduling optimization solver; The large model is called to instruct the large model to obtain a device execution plan that matches the user query based on the optimal solution of the energy storage decision, and the device execution plan is output as an output result. The device execution plan is an adjustment plan and execution instructions for the user's energy storage during the power adjustable period.

3. The method according to claim 1 or 2, characterized in that: The time series model is pre-trained using power market transaction data as training samples and a set of future power market electricity prices at time nodes corresponding to the power market transaction data as training labels.

4. The method according to claim 1 or 2, characterized in that: The process of determining the adjustable power period and the price difference of the adjustable power period according to the short-term power market price and the relatively long-term power market price includes: When the difference between the short-term electricity market price and the relatively long-term electricity market price is greater than a preset threshold, the difference is used as the adjustable period price difference, and the time node interval corresponding to the time when the difference between the short-term electricity market price and the relatively long-term electricity market price is greater than the preset threshold is used as the adjustable electricity period.

5. The method according to claim 1 or 2, characterized in that: The short-term electricity market price is the real-time electricity market price, and correspondingly, the relatively long-term electricity market price is the day-ahead electricity market price or the medium- and long-term electricity market price; or, The short-term electricity market price is the day-ahead electricity price in the electricity market, and correspondingly, the relatively long-term electricity market price is the medium- and long-term electricity price in the electricity market.

6. The method according to claim 2, characterized in that The execution actions of the training process of the reinforcement learning model include not only the charging and discharging power of the user's energy storage, but also constraints, which include: The capacity of the user energy storage cannot be lower than the minimum capacity of the user energy storage or exceed the maximum capacity of the user energy storage, the power market load demand cannot exceed the charging and discharging capacity of the user energy storage, the user energy storage cannot be charged and discharged at the same time, and the net change of the capacity of the user energy storage plus charging and discharging cannot exceed the maximum capacity of the user energy storage.

7. The method according to claim 1 or 2, characterized in that: The power market transaction data includes: power market load demand, power market controllable load information, and / or power grid operation status information.

8. The method according to claim 2, characterized in that: The reinforcement learning model is a Q-learning model, a proximal policy optimization PPO model or a deep Q network DQN model.

9. A virtual power plant dispatching device based on a large model, characterized in that: include: A first instruction response unit is used to obtain a first query instruction of a user through a large model, and analyze the query intention of the first query instruction. If the query intention is to query the adjustable power period and its price difference, then obtain power market transaction data; A time series model calling unit, configured to call a pre-trained time series model through the large model, so as to predict a future power market price set based on the power market transaction data through the time series model, wherein the set includes a short-term power market price and a relatively long-term power market price; A data analysis unit, configured to determine an adjustable power period and a price difference of the adjustable power period according to the short-term power market price and the relatively long-term power market price; The first intention result output unit is used to output the adjustable power period and the price difference of the adjustable power period as output results.

10. An electronic device, characterized in that: include: Memory and processor; The memory is used to store programs; The processor is used to execute the program to implement each step of the virtual power plant scheduling method as described in any one of claims 1 to 8.

Images