Virtual power plant scheduling method and related equipment

Abstract

The invention provides a scheduling method of a virtual power plant and related equipment. Specifically, the scheduling method comprises the following steps: acquiring price constraint conditions, load demand data and state data of the power generation units; according to the state data and the load demand data, determining first output power of each power generation unit by using a trained deep Q network; according to the first output power and a price constraint condition, determining second output power of each power generation unit by using a trained deep Q network; wherein the deep Q network comprises a first deep Q network and a second deep Q network; wherein the first deep Q network is configured to determine a first output power; the second deep Q network is configured to determine a second output power. The deep Q network is used to fully analyze the load demand data, the operation state of the power generation unit and the price constraint condition, thereby facilitating the reduction of the load loss, and promoting the application and development of the virtual power plant in the energy field.

Description

Technical Field

[0001] The present disclosure relates to the field of power grid technology, and in particular to a scheduling method and related equipment for a virtual power plant. Background Art

[0002] In the application of virtual power plants (VPPs), in order to reduce the operating costs of power grids, achieve sustainable use of energy, and thus improve the application performance of virtual power plants, the scheduling problem has received widespread attention in the field. The scheduling methods of virtual power plants provided by related technologies have drawbacks in practical applications, resulting in scheduling results that are difficult to meet the operational needs of virtual power plants. Summary of the invention

[0003] In view of this, the purpose of the present disclosure is to propose a scheduling method and related equipment for a virtual power plant.

[0004] Based on the above purpose, the present disclosure provides a scheduling method for a virtual power plant, wherein the virtual power plant includes at least one power generation unit; the scheduling method specifically includes: obtaining price constraints, load demand data, and status data of the power generation unit; Determine the first output power of each of the power generation units using a trained deep Q network according to the state data and the load demand data; According to the first output power and the price constraint, the second output power of each of the power generation units is determined using a trained deep Q network; wherein, The deep Q network includes a first deep Q network and a second deep Q network; wherein the first deep Q network is configured to determine the first output power; and the second deep Q network is configured to determine the second output power.

[0005] Based on the same inventive concept, an embodiment of the present disclosure further provides a scheduling device for a virtual power plant, wherein the virtual power plant includes at least one power generation unit; the scheduling device includes: An acquisition module is configured to: acquire price constraints, load demand data and status data of the power generation unit; A first output power module is configured to: determine the first output power of each of the power generation units according to the state data and the load demand data using a trained deep Q network; The second output power module is configured to: determine the second output power of each of the power generation units according to the first output power and the price constraint condition using a trained deep Q network; wherein, The deep Q network includes a first deep Q network and a second deep Q network; wherein the first deep Q network is configured to determine the first output power; and the second deep Q network is configured to determine the second output power.

[0006] Based on the same inventive concept, an embodiment of the present disclosure also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a scheduling method as described in any one of the above is implemented.

[0007] Based on the same inventive concept, an embodiment of the present disclosure further provides a computer program product, including computer program instructions. When the computer program instructions are executed on a computer, the computer executes any of the above-mentioned scheduling methods.

[0008] From the above, it can be seen that the present disclosure provides a virtual power plant scheduling method and related equipment, by pre-constructing a deep Q network including a first deep Q network and a second deep Q network, according to the state data and the load demand data, using the trained first deep Q network to determine the first output power of each power generation unit; and then according to the first output power and the price constraint, using the second deep Q network to determine the second output power of each power generation unit. With such a technical solution, the deep Q network is used to fully analyze the load demand data, the operating status of the power generation unit and the price constraint, which helps to reduce load losses and promote the application and development of virtual power plants in the energy field. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the present disclosure or related technologies, the drawings required for use in the embodiments or related technical descriptions are briefly introduced below. Obviously, the drawings described below are only embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A schematic diagram of the structure of a deep Q network provided by an embodiment of the present disclosure is shown; Figure 2 A schematic diagram of a topological structure of a virtual power plant operation provided by an embodiment of the present disclosure is shown; Figure 3 A schematic diagram of a power spot market system structure provided by an embodiment of the present disclosure is shown; Figure 4 A schematic diagram showing a mechanism of operation of a power trading market provided by an embodiment of the present disclosure; Figure 5 A schematic diagram showing a result of upper layer optimization scheduling provided by an embodiment of the present disclosure is shown; Figure 6 A schematic diagram showing a result of lower layer optimization scheduling provided by an embodiment of the present disclosure is shown; Figure 7 A schematic diagram showing a comparison of virtual power plant scheduling results provided by an embodiment of the present disclosure is shown; Figure 8 A schematic diagram showing a flow chart of a scheduling method for a virtual power plant provided by an embodiment of the present disclosure; Fig. 9 A schematic diagram showing the structure of a scheduling device for a virtual power plant provided by an embodiment of the present disclosure is shown; Fig.10 A schematic structural diagram of an electronic device provided by an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0011] In order to make the objectives, technical solutions and advantages of the present disclosure more clearly understood, the present disclosure is further described in detail below in combination with specific embodiments and with reference to the accompanying drawings.

[0012] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the embodiments of the present disclosure should be understood by people with ordinary skills in the field to which the present disclosure belongs. The "first", "second" and similar words used in the embodiments of the present disclosure do not indicate any order, quantity or importance, but are only used to distinguish different components. "Including" or "comprising" and similar words mean that the elements or objects appearing before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0013] In order to facilitate understanding of the technical solutions of the present disclosure, some technical terms involved in the present disclosure are introduced below.

[0014] Virtual power plant is an IoT technology that aggregates adjustable loads, energy storage and power supply resources in different spaces to achieve autonomous coordination and optimization control, and aggregate participation in power system operation and power market transactions. Virtual power plant mainly consists of three parts: power generation system, energy storage equipment and communication system.

[0015] An agent is an entity that can act autonomously and has a certain decision-making ability in the system. In the virtual power plant environment, each distributed generation unit can be regarded as an agent, which can autonomously adjust its power generation output according to the environmental status and scheduling strategy.

[0016] Deep Q-Network (DQN) is a reinforcement learning algorithm that combines Q learning and deep learning. It approximates the Q function through a neural network to achieve efficient decision-making in complex environments.

[0017] As mentioned in the background technology section, the scheduling problem of virtual power plants has received widespread attention in the field. For example, Cheng Xueting et al. proposed a two-layer optimization scheduling method for virtual power plants considering the uncertainty risk of renewable energy output (see reference 1, Cheng Xueting, Bao Yueshuang, Jin Yulong, et al. Two-layer scheduling strategy for virtual power plants considering the uncertainty risk of renewable energy output [J]. Modern Electric Power, 2023, 40 (06): 967-975.). This scheduling method analyzes the uncertainty of renewable energy output under different parameter conditions and quantifies the corresponding risk value; then, combined with the internal resource status of the virtual power plant, a two-layer optimization scheduling model is constructed; finally, the optimal scheduling strategy is obtained by solving the model. This method is highly dependent on the energy storage equipment in the virtual power plant. However, the capacity and charging and discharging efficiency of the energy storage equipment are limited, which may limit the effectiveness of the two-layer optimization scheduling model. For example, Pan Ting et al. proposed a two-layer optimization scheduling method for virtual power plants based on dynamic time-of-use electricity prices (see reference 2, Pan Ting, Dong Houqi, Wang Yuqing, et al. Research on two-layer optimization scheduling of virtual power plants based on dynamic time-of-use electricity prices [J]. Systems Science and Mathematics, 2024, 44 (02): 304-325.). This scheduling method first constructs a two-layer optimization scheduling model for virtual power plants with power consumption units connected, then analyzes the load side demand of the power plant at the upper level of the model, analyzes the dynamic time-of-use electricity price at the lower level, and establishes an objective function with the minimum operating cost, thereby solving the optimal scheduling result. However, this method does not consider the impact of new energy on the scheduling process, which may reduce the accuracy of the model solution. For example, Zhao Lihang et al. proposed a two-layer optimization dispatch method for virtual power plants based on energy economics (see reference 3, Zhao Lihang, Chang Weiguang, Yang Min, et al. Two-stage energy economic optimization dispatch of virtual power plants under power market environment [J]. China Electric Power, 2022, 55 (10): 14-22.). This dispatch method first constructs a two-layer optimization dispatch model of day-ahead and intraday; using the day-ahead forecast data as a reference, the net load parameters are adjusted for the intraday dispatch layer, and the dispatch result is obtained by solving the energy economics as the objective function. However, in some cases, excessive pursuit of economic benefits may lead to a decrease in the stability of the model solution results.

[0018] Taking into account the drawbacks of the virtual power plant two-layer optimization scheduling method proposed in the related art in actual application, which makes it difficult for the scheduling results to meet the operational needs of the virtual power plant, the embodiment of the present disclosure provides a scheduling method and related equipment for a virtual power plant, by pre-constructing a deep Q network including a first deep Q network and a second deep Q network, and determining the first output power of each power generation unit according to the state data and the load demand data using the trained first deep Q network; and then determining the second output power of each power generation unit according to the first output power and the price constraint using the second deep Q network. With such a technical solution, the deep Q network is used to fully analyze the load demand data, the operating status of the power generation unit and the price constraint, which helps to reduce load losses and promote the application and development of virtual power plants in the energy field.

[0019] In order to make the technical solution of the present disclosure clearer and easier to understand, a scheduling method for a virtual power plant provided by an embodiment of the present disclosure is introduced below with reference to the accompanying drawings.

[0020] In order to achieve unified scheduling of distributed power generation units in a virtual power plant, the embodiment of the present disclosure introduces a deep Q network to establish a unified scheduling node state space. It should be noted that the distributed power generation units may include one or more of photovoltaic units that generate electricity using photovoltaic panels, wind power units that generate electricity using wind turbines, and thermal power units that generate electricity using gas, etc., and the present disclosure does not limit this. In addition, for a virtual power plant, the number of photovoltaic units, wind power units, and thermal power units can be zero, one, or multiple.

[0021] First, define the virtual power plant dispatch node set. The dispatch node set represents all distributed generation units that need to be dispatched in the virtual power plant, as shown in formula (1): (1) in, V represents the set of scheduling nodes, V 1 represents the i-th distributed generation unit, n Indicates the number of schedulable nodes.

[0022] Next, the deep Q network algorithm is used to assign a corresponding agent to each node in the scheduling node set. It should be noted that the state representation of the agent includes but is not limited to the current power generation output, load demand, uncertainty parameters of wind power and photovoltaic power, etc. of the scheduling node. For example, the load demand state represents the power demand that the virtual power plant needs to meet in different time periods, which is usually related to the user's electricity satisfaction. For example, the output state of the power generation unit: including the actual or predicted output of renewable energy power generation units such as wind power and photovoltaic power, which are affected by natural conditions such as wind speed and light intensity; for example, the power market price state: represents the transaction price in different time periods in the power market, which will affect the purchase and sale decisions of the virtual power plant; for example, the energy storage device state includes the current power of the energy storage device, the charging and discharging efficiency, the capacity limit, etc., which affect the virtual power plant's ability to adjust when the power supply and demand are unbalanced; for example, the environmental parameter state: such as environmental factors such as temperature and humidity, which may affect the efficiency and output of the power generation unit; for example: the equipment operation state: including the health status, maintenance plan, failure rate, etc. of the power generation equipment, which is crucial for predicting power generation capacity and planning maintenance activities. The action space corresponds to the operations of adjusting power generation output, energy storage charging and discharging, etc. of the dispatching node. By setting a reasonable reward function, the intelligent agent can maximize long-term benefits (such as minimizing load loss rate, reducing operating costs, etc.) during the optimization scheduling process.

[0023] Then, define the set of agents, namely: A ={ A 1 ,A 2 ,…，A n};in, A represents the set of agents, A i represents the i-th agent, n Represents the number of agents.

[0024] Next, with the agent as the core, all states of the scheduling node are merged to generate the state space matrix of the scheduling node, as shown in formula (2): (2) Where S represents the state space matrix of the scheduling node, m Represents the agent dimension.

[0025] Add the executable actions of the scheduling node to the state space matrix. The space matrix of the executable action is designed to be a pair matrix with the state space matrix. According to the matrix joint relationship, the weight of the scheduling node in the deep Q network is calculated, as shown in formula (3): (3) in, represents the scheduling node weight, Indicates executable actions. represents the reward function of the deep Q-network.

[0026] In some embodiments, the corresponding reward function is defined by setting the exploration decay function of the algorithm, as shown in formula (4): (4) In the formula, represents the exploration decay function, represents the minimum exploration tolerance, represents a natural constant, Indicates time.

[0027] A deep Q network with two layers of convolution is constructed in the state space with defined node weights. Its architecture is as follows Figure 1 shown.

[0028] like Figure 1 As shown, the double-layer convolutional deep Q network provided by the embodiment of the present disclosure includes a first network and a second network. Among them, the first network includes a first convolutional layer and a first fully connected layer; the second network includes a second convolutional layer and a second fully connected layer. The first convolutional layer and the second convolutional layer are respectively connected to the state and action of the scheduling node, and the execution of the double-layer optimization scheduling strategy in the state space of the scheduling node is realized through the connection of the action parameters. Here, the first network can predict the Q value based on the current state and action, and update its parameters by the gradient descent method.

[0029] In some embodiments, the deep Q network also includes a loss function for fully connected layer regression, as shown in formula (5): (5) in, L represents the loss function, d represents the window size of the fully connected layer, w 0 represents the weight of the fully connected layer, P(j) Represents neighbor nodes j The random walk probability of R represents the regression coefficient.

[0030] During training, the network updates its parameters by backpropagating through the loss function (the difference calculated based on the reward function).

[0031] Through the above steps, the initial deep Q network is constructed.

[0032] Based on the state space of the scheduling node and the deep Q network, a two-layer virtual power plant optimization scheduling architecture is designed and connected to at least two convolutional layers respectively. It should be noted that the two-layer scheduling of the virtual power plant includes upper-layer optimization scheduling and lower-layer optimization scheduling.

[0033] In the upper-level optimization scheduling architecture, the present disclosure analyzes the load demand of the virtual power plant and the output uncertainty of wind power and photovoltaic power respectively to preliminarily generate the power parameters of the scheduling node.

[0034] Based on the historical data of the virtual power plant, its load demand parameters per unit time are calculated, as shown in formula (6): (6) in, E represents the load demand parameter, P(t) express t Total power of the virtual power plant at time, Represents a unit time period, Indicates the user's electricity satisfaction at the corresponding time, and They represent the load power consumption before and after demand response respectively.

[0035] Combined with the output uncertainty of the wind power and photovoltaic power generation units in the virtual power plant, the corresponding output power characteristics are calculated.

[0036] In some embodiments, the embodiments of the present disclosure define the output power characteristics of wind power uncertainty output according to the real-time wind speed, as shown in formula (7): (7) in, represents the uncertainty power of wind power, It indicates the output power of the wind turbine at rated wind speed. represents the shape parameter, represents the scale parameter, Indicates the real-time wind speed of the unit environment.

[0037] In some embodiments, the output power characteristic of the photovoltaic uncertainty output is defined according to the real-time illumination, as shown in formula (8): (8) in, represents the uncertainty of the photoelectric power, Indicates the rated output power of the photovoltaic module. Indicates the power derating factor, represents the standard illumination, represents the photovoltaic array capacity, represents the photovoltaic panel temperature under standard conditions, Indicates the real-time temperature of the solar panel. represents the temperature coefficient, Indicates the real-time light intensity.

[0038] Combining the above three parameters, the upper-layer optimization scheduling strategy is generated, and the corresponding expression can be shown as formula (9): (9) in, Indicates the node power parameter for upper layer optimization scheduling.

[0039] Finally, on the basis of the upper-layer optimization scheduling strategy, the power trading market environment of the virtual power plant is analyzed at the lower layer to minimize the net load fluctuation, thereby further correcting the power parameters of the scheduling node based on the upper-layer optimization scheduling strategy.

[0040] In some embodiments, a Boolean variable is introduced into the power market trading system to analyze the price constraints of virtual power plant scheduling, as shown in formula (10): (10) in, H represents the electricity trading volume, Indicates the highest transaction volume, represents the electricity transaction price, Represents a Boolean variable.

[0041] Under this constraint, the net load correction value of the power parameter obtained by the upper-layer optimization scheduling is calculated as shown in formula (11): (11) in, Indicates the net load correction value.

[0042] The calculated net load correction value is coupled with the power parameter of the upper-level optimization scheduling to obtain the power parameter of the lower-level optimization scheduling, as shown in formula (12): (12) In the formula, Indicates the power parameter of the lower layer optimization scheduling.

[0043] Substitute the optimization scheduling results of the upper layer and the lower layer into the action unit in the state space of the scheduling node to implement the execution of the scheduling strategy. Here, the power parameters of the upper layer optimization scheduling and the power parameters of the lower layer optimization scheduling can be used as limiting conditions to adjust the scheduling strategy. , etc. can be used as restriction conditions for the first network; illustratively, power parameters and price constraints of the lower-layer optimization scheduling can be used as restriction conditions for the second network.

[0044] Below, the embodiments of the present disclosure provide a detailed description of the scheduling method of the virtual power plant described above in combination with the actual operation architecture of a virtual power plant.

[0045] Based on the actual operation architecture of a virtual power plant, the corresponding topology is constructed, such as Figure 2 As shown. Figure 2 As shown in the figure, the constructed virtual power plant includes three distributed power generation units: photovoltaic, wind power and gas. The virtual power plant of this architecture is integrated into the power spot market system consistent with the actual situation of a certain place. The market system is used to establish the day-ahead market environment of the virtual power plant. Its system architecture is as follows: Figure 3 shown.

[0046] like Figure 3 In the market system shown in FIG, the actual market data collected is used as the actual power trading environment of the virtual power plant. In the experiment of this embodiment, the trading phase of the virtual power plant in the power spot market is set to one hour. This generates the power trading market operation mechanism in this experiment, such as Figure 4 shown.

[0047] The operating parameters of each power generation unit are set for the virtual power plant operation system constructed above, as shown in Table 1.

[0048] Table 1 Operating parameters of virtual power plant power generation units

[0049] Based on the above power generation unit parameter settings, the fixed energy consumption of the power-to-gas equipment of the experimental virtual power plant is defined as 0.179MW.

[0050] Based on the settings of the above parameters, the scheduling test of this virtual power plant was carried out in the constructed virtual power plant environment.

[0051] According to the virtual power plant dispatching environment set in this experiment, the parameters of the deep Q network in the dispatching method of the embodiment of the present disclosure are set, as shown in Table 2.

[0052] Table 2 Deep Q network parameter settings

[0053] The training of the deep Q network provided by the embodiment of the present disclosure is completed according to the parameters in Table 2. On this basis, the scheduling method provided by the embodiment of the present disclosure is used to perform scheduling tests on the virtual power plant in this environment.

[0054] After analyzing the scheduling method provided by the embodiment of the present disclosure, the optimization scheduling results of the upper layer and the lower layer are obtained respectively, such as Figure 5 , 6 shown.

[0055] according to Figure 5 and Figure 6 It can be seen that in the upper-level scheduling, the power of each unit is preliminarily designed according to the load demand of the virtual power plant and the output uncertainty of the power generation unit. In the lower-level scheduling, the net load fluctuation is minimized under the constructed power trading market operation mechanism, and the energy storage limitations in the virtual power plant are fully considered to obtain the final optimized scheduling result. Analysis of the two-layer optimization scheduling results shows that the scheduling method of the embodiment of the present disclosure, on the basis of meeting the operating requirements and limitations of the virtual power plant, makes full use of the high-efficiency operating periods of wind power and photovoltaic power, increases the output in the time periods of 6~10 and 10~16 respectively, and increases the output of the gas turbine in the remaining time periods. It can be seen that the scheduling result obtained by the scheduling method of the embodiment of the present disclosure can make full use of the output characteristics of the virtual power plant and has a certain feasibility.

[0056] In order to further verify the practical application performance of the scheduling method of the embodiment of the present disclosure, the virtual power plant double-layer scheduling methods proposed in References 1, 2 and 3 are respectively used to carry out comparative experiments. The duration of each test round is set to 48 hours. The scheduling effectiveness of each method is analyzed by calculating the load loss rate of each method during this period. The scheduling results are as follows: Figure 7 shown.

[0057] like Figure 7 As shown, the load loss rate of the methods proposed in Document 1 and Document 3 is relatively high, indicating that there is a large waste of resources in their scheduling results. The method in Document 2 and the scheduling method provided in the embodiment of the present disclosure show a smaller load loss rate, but a comparison shows that the method of the present disclosure is more stable. It can be seen that the scheduling method of the present disclosure not only has a better scheduling effect, but is also relatively stable, and has a higher practical application value in the operation of virtual power plants.

[0058] Based on the same inventive concept, the embodiment of the present disclosure also provides a scheduling method for a virtual power plant. Here, the virtual power plant includes at least one power generation unit, such as a wind power generation unit, a photovoltaic power generation unit, and the like. Figure 8 A flow chart of a virtual power plant scheduling method provided by an embodiment of the present disclosure is shown; Figure 8 As shown, the scheduling method includes: S801: Obtain price constraints, load demand data, and status data of the power generation unit; here, the price constraints can refer to formula (10) and are not repeated here; the status data may include environmental data (such as real-time wind speed, real-time light intensity, real-time temperature, etc.), operating data (such as health status, failure rate, rated power, tip speed ratio, etc.), etc.

[0059] S803: Determine the first output power of each of the power generation units according to the state data and the load demand data using a trained deep Q network; here, the first output power of each of the power generation units can be as follows: Figure 5 As shown; S805: Determine the second output power of each of the power generation units according to the first output power and the price constraint condition using a trained deep Q network; here, the second output power of each of the power generation units can be as follows: Figure 6 As shown; among them, The deep Q network includes a first deep Q network (e.g., a first network) and a second deep Q network (e.g., a second network); wherein the first deep Q network is configured to determine the first output power; and the second deep Q network is configured to determine the second output power.

[0060] Adopting such a technical solution and using the deep Q network to fully analyze the load demand data, the status data of the power generation unit and the price constraints will help reduce load losses and promote the application and development of virtual power plants in the energy field.

[0061] In some embodiments, the step of obtaining load demand data specifically includes: Obtain the historical data of the virtual power plant; here, the historical data may include the total power of the virtual power plant P(t) , User satisfaction with electricity consumption , and wait.

[0062] Referring to formula (6), the load demand data per unit time is determined according to the historical data. Here, the unit time can be preset, such as 1 hour, 2 hours, 4 hours, 12 hours, etc., which is not limited in the present disclosure.

[0063] In some embodiments, reference is made to Figure 1 , the first deep Q network includes at least one first convolutional layer and a first fully connected layer; the second deep Q network includes at least one second convolutional layer and a second fully connected layer; wherein a regression loss function is configured between the first fully connected layer and the second fully connected layer. Here, the regression loss function can be shown as formula (5).

[0064] In some embodiments, the at least one power generation unit includes at least one of a wind power unit and a photovoltaic unit; As shown in reference formula (7), the restriction condition of the first output power (or the second output power) of the wind power unit includes not being higher than the first power (corresponding to the uncertain power) of the wind power unit; the first power of the wind power unit is determined based on the state data of the wind power unit; the state data at least includes the real-time wind speed; and / or As shown in reference formula (8), the restriction condition of the first output power (or the second output power) of the photoelectric unit includes that it is not higher than the second power (corresponding to the uncertain power) of the photoelectric unit; the second power of the photoelectric unit is determined based on the status data of the photoelectric unit; and the status data at least includes the real-time light intensity.

[0065] Such a technical solution can ensure that the uncertain power of the wind power unit and the uncertain power of the photovoltaic unit participate in the scheduling of the virtual power plant, schedule the first output power and the second output power within the maximum range of the uncertain power, and give full play to the role of the wind power unit and the photovoltaic unit.

[0066] In some embodiments, the constraint of the second deep Q network includes a price constraint; wherein, referring to formula (10) and formula (11), the price constraint is used to modify the first output power. In the second deep Q network, considering the price constraint is conducive to minimizing the net load fluctuation based on a full analysis of the market price.

[0067] In some embodiments, forming the deep Q network specifically includes: Based on each power generation unit of the virtual power plant, construct a node state space matrix (such as formula (2)) and a node weight (such as formula 3); Based on the weights and the node state space matrix, construct the original deep Q network (reference Figure 1 ); Obtain network parameters (such as Table 2) and a training set, and train the original deep Q network based on the network parameters and the training set to obtain the deep Q network.

[0068] In some embodiments, constructing a node state space matrix and a node weight based on each power generation unit of the virtual power plant includes: Taking each power generation unit of the virtual power plant as a node, a node set is constructed (e.g., formula (1)); Based on the node set, construct an agent set; Based on the agent set, the node state space matrix (eg, formula (2)) is constructed. It should be noted that the above steps can refer to the above and will not be repeated here.

[0069] It should be noted that the method of the embodiment of the present disclosure can be performed by a single device, such as a computer or a server. The method of the present embodiment can also be applied in a distributed scenario and completed by multiple devices cooperating with each other. In the case of such a distributed scenario, one of the multiple devices can only perform one or more steps in the method of the embodiment of the present disclosure, and the multiple devices will interact with each other to complete the described method.

[0070] It should be noted that the above describes some embodiments of the present disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the above embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0071] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present disclosure also provides a scheduling device for a virtual power plant.

[0072] refer to Fig. 9 , the scheduling device comprises: The acquisition module 901 is configured to: acquire price constraints, load demand data and status data of the power generation unit; The first output power module 903 is configured to: determine the first output power of each of the power generation units according to the state data and the load demand data using a trained deep Q network; The second output power module 905 is configured to: determine the second output power of each of the power generation units according to the first output power and the price constraint condition using a trained deep Q network; wherein, The deep Q network includes a first deep Q network and a second deep Q network; wherein the first deep Q network is configured to determine the first output power; and the second deep Q network is configured to determine the second output power.

[0073] It should be noted that the deep Q network can be constructed and trained through network modules, which will not be described in detail.

[0074] For the convenience of description, the above device is described by dividing it into various modules according to its functions. Of course, when implementing the present disclosure, the functions of each module can be implemented in the same or multiple software and / or hardware.

[0075] The device of the above embodiment is used to implement the corresponding scheduling method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.

[0076] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments and methods, the present disclosure also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the scheduling method described in any of the above embodiments is implemented.

[0077] Fig.10 A more specific schematic diagram of the hardware structure of an electronic device provided in this embodiment is shown, and the device may include: a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040 are connected to each other through the bus 1050 in the device.

[0078] The processor 1010 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0079] The memory 1020 may be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 may store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program codes are stored in the memory 1020 and are called and executed by the processor 1010.

[0080] The input / output interface 1030 is used to connect the input / output module to realize information input and output. The input / output module can be configured in the device as a component (not shown in the figure), or it can be externally connected to the device to provide corresponding functions. The input device may include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device may include a display, a speaker, a vibrator, an indicator light, etc.

[0081] The communication interface 1040 is used to connect a communication module (not shown) to realize communication interaction between the device and other devices. The communication module can realize communication through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0082] The bus 1050 includes a path that transmits information between the various components of the device (eg, the processor 1010 , the memory 1020 , the input / output interface 1030 , and the communication interface 1040 ).

[0083] It should be noted that, although the above device only shows the processor 1010, the memory 1020, the input / output interface 1030, the communication interface 1040 and the bus 1050, in the specific implementation process, the device may also include other components necessary for normal operation. In addition, it can be understood by those skilled in the art that the above device may also only include the components necessary for implementing the embodiments of the present specification, and does not necessarily include all the components shown in the figure.

[0084] The electronic device of the above embodiment is used to implement the corresponding scheduling method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.

[0085] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present disclosure also provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the scheduling method described in any of the above embodiments.

[0086] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.

[0087] The computer instructions stored in the storage medium of the above embodiments are used to enable the computer to execute the scheduling method described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0088] Based on the same inventive concept, corresponding to the scheduling method described in any of the above embodiments, the present disclosure also provides a computer program product, which includes computer program instructions. In some embodiments, the computer program instructions can be executed by one or more processors of a computer so that the computer and / or the processor executes the scheduling method. Corresponding to the execution subject corresponding to each step in each embodiment of the scheduling method, the processor that executes the corresponding step may belong to the corresponding execution subject.

[0089] The computer program product of the above embodiment is used to enable the computer and / or the processor to execute the scheduling method described in any of the above embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0090] It should be understood by those skilled in the art that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present disclosure (including the claims) is limited to these examples; based on the concept of the present disclosure, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of different aspects of the embodiments of the present disclosure as described above, which are not provided in detail for the sake of simplicity.

[0091] In addition, to simplify the description and discussion, and in order not to obscure the embodiments of the present disclosure, known power / ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided figures. In addition, devices may be shown in the form of block diagrams to avoid obscuring the embodiments of the present disclosure, and this also takes into account the fact that the details of the implementation of these block diagram devices are highly dependent on the platform on which the embodiments of the present disclosure are to be implemented (i.e., these details should be fully within the scope of understanding of those skilled in the art). Where specific details (e.g., circuits) are set forth to describe exemplary embodiments of the present disclosure, it is apparent to those skilled in the art that the embodiments of the present disclosure may be implemented without these specific details or with variations in these specific details. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0092] Although the present disclosure has been described in conjunction with specific embodiments of the present disclosure, many alternatives, modifications and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may use the discussed embodiments.

[0093] The embodiments of the present disclosure are intended to cover all such substitutions, modifications and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present disclosure should be included in the scope of protection of the present disclosure.

Claims

1. A scheduling method for a virtual power plant, characterized in that: The virtual power plant includes at least one power generation unit; the scheduling method includes: Obtaining price constraints, load demand data, and status data of the power generation unit; Determine the first output power of each of the power generation units using a trained deep Q network according to the state data and the load demand data; According to the first output power and the price constraint, the second output power of each of the power generation units is determined using a trained deep Q network; wherein, The deep Q network includes a first deep Q network and a second deep Q network; wherein the first deep Q network is configured to determine the first output power; and the second deep Q network is configured to determine the second output power.

2. The scheduling method according to claim 1, characterized in that: The steps for obtaining load demand data include: Obtaining historical data of the virtual power plant; The load demand data per unit time is determined according to the historical data.

3. The scheduling method according to claim 1, characterized in that: The first deep Q network includes at least one first convolutional layer and a first fully connected layer; the second deep Q network includes at least one second convolutional layer and a second fully connected layer.

4. The scheduling method according to claim 1, characterized in that: The at least one power generation unit includes at least one of a wind power unit and a photovoltaic unit; The limiting condition of the first output power of the wind power unit includes not being higher than the first power of the wind power unit; the first power of the wind power unit is determined based on the status data of the wind power unit; the status data at least includes the real-time wind speed; and / or The limiting condition of the first output power of the photoelectric unit includes not being higher than the second power of the photoelectric unit; the second power of the photoelectric unit is determined based on the status data of the photoelectric unit; the status data at least includes the real-time light intensity.

5. The scheduling method according to claim 1, characterized in that: The restriction condition of the second deep Q network includes a price constraint condition; wherein the price constraint condition is used to modify the first output power.

6. The scheduling method according to claim 1, characterized in that: Forming the deep Q network specifically includes: Based on each power generation unit of the virtual power plant, construct a node state space matrix and node weights; Constructing an original deep Q network based on the weights and the node state space matrix; Obtain network parameters and a training set, and train the original deep Q network based on the network parameters and the training set to obtain the deep Q network.

7. The scheduling method according to claim 6, characterized in that: The step of constructing a node state space matrix and a node weight based on each power generation unit of the virtual power plant includes: Taking each power generation unit of the virtual power plant as a node, constructing a node set; Based on the node set, construct an agent set; Based on the agent set, the node state space matrix is ​​constructed.

8. A virtual power plant dispatching device, characterized in that: The virtual power plant includes at least one power generation unit; The scheduling device comprises: An acquisition module is configured to: acquire price constraints, load demand data and status data of the power generation unit; A first output power module is configured to: determine the first output power of each of the power generation units according to the state data and the load demand data using a trained deep Q network; The second output power module is configured to: determine the second output power of each of the power generation units according to the first output power and the price constraint condition using a trained deep Q network; wherein, The deep Q network includes a first deep Q network and a second deep Q network; wherein the first deep Q network is configured to determine the first output power; and the second deep Q network is configured to determine the second output power.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable by the processor, characterized in that: The processor implements the scheduling method according to any one of claims 1 to 7 when executing the computer program.

10. A computer program product, characterized in that The method comprises computer program instructions, and when the computer program instructions are executed on a computer, the computer is caused to execute the scheduling method according to any one of claims 1 to 7.

Images