Power information processing method and system based on distributed energy storage
By collaboratively optimizing the power distribution of distributed energy storage devices through LSTM networks and deep reinforcement learning algorithms, the problem of insufficient utilization of the remaining energy storage capacity of distributed energy storage devices is solved, and efficient management of power resources and stable operation of microgrids are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- TOYOTA JIDOSHA KK
- Filing Date
- 2021-07-13
- Publication Date
- 2026-06-19
AI Technical Summary
The remaining energy storage capacity of distributed energy storage devices is difficult to utilize effectively, resulting in a waste of power resources, and it is difficult to achieve aggregate management when the number of end users is large.
An LSTM network is used for power load and supply forecasting, combined with a deep reinforcement learning algorithm to construct a power disposal scheme, dynamically manage the energy storage capacity pool, and coordinate power allocation between the generation and consumption sides.
It improves the residual regulation capacity of distributed energy storage, ensures the safe and stable operation of the system, maximizes the benefits of microgrids, and avoids power waste.
Smart Images
Figure CN115622099B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of distributed energy intelligent trading, and in particular to a power information processing method and system based on distributed energy storage. Background Technology
[0002] Currently, many countries and regions are constructing energy storage systems on a large scale. With the rapid decline in energy storage costs, distributed energy storage is gradually becoming more widespread. Unlike centralized energy storage facilities, distributed energy storage devices are owned by the respective end users. For example, end users on the power supply side can use distributed energy storage to optimize their electricity consumption and reduce electricity costs. However, the load demand of a single energy storage device is limited, thus limiting energy optimization. For instance, the remaining energy storage capacity of distributed energy storage devices may not be effectively utilized, resulting in waste. Furthermore, given the large number of end users, how to aggregate and process the remaining energy storage capacity of a large number of distributed energy storage devices without causing a huge waste of electricity resources is an urgent problem to be solved. Summary of the Invention
[0003] The purpose of this application is to provide a power information processing method and system based on distributed energy storage. This method can effectively manage a large number of power supply terminals that are distributed energy storage, and avoid the waste caused by the power supply terminals having excess energy storage capacity because they cannot distribute all the power they generate.
[0004] To address the aforementioned technical problems, embodiments of this application employ the following technical solution: a power information processing method based on distributed energy storage, comprising:
[0005] The power load and power supply during the power handling period are predicted by the LSTM network, and the corresponding prediction results are generated. The power supply corresponds to multiple power supply terminals, which are set up in a distributed manner.
[0006] Based on the prediction results, the required backup energy storage capacity for each unit time period within the power handling time period is determined, wherein the backup energy storage capacity is associated with the power load;
[0007] Publish energy storage demand characterizing the backup energy storage capacity, and obtain feedback information based on the published energy storage demand;
[0008] Under the first constraint, a corresponding power disposal scheme is constructed based on the energy storage demand and the feedback information;
[0009] Based on the constructed power disposal scheme, a deep reinforcement learning algorithm is used to control the power supply end for power supply under the second constraint to implement power disposal.
[0010] Optionally, before controlling the power supply end for power supply to implement power disposal, the method further includes:
[0011] Obtain information for adjusting the power disposal plan;
[0012] Under the third constraint, the power disposal plan is adjusted based on the adjustment information.
[0013] Optionally, adjusting the power disposal plan based on the adjustment information under the third constraint includes:
[0014] At least the supply quantity corresponding to the power supply terminal and the processing quantity of the supply quantity are subject to a first constraint operation;
[0015] If the first constraint operation meets the third constraint condition, a first objective function is constructed based on the disposal information and disposal results formed during the power transfer process, wherein the first objective function takes increasing the amount of power disposal as the calculation objective.
[0016] The power disposal scheme is adjusted based on the first objective function.
[0017] As an option, where,
[0018] The first objective function is:
[0019]
[0020] The expression corresponding to the third constraint condition is:
[0021]
[0022] Where t is time, and These refer to the import and export quotas generated during the power disposal process. and These represent whether power-related actions based on the introduced and output quotas have been implemented, respectively. A value of 1 indicates implementation, while a value of 0 indicates non-implementation. For the electricity disposal quota, The supply quantity corresponding to the supply end i. The disposal quantity corresponding to supply end j. This indicates a constraint operation.
[0023] Optionally, under the first constraint, constructing a corresponding power disposal scheme based on the energy storage demand and the feedback information includes:
[0024] A second constraint operation is performed on the power supply quantity corresponding to the power supply end and the power disposal quota corresponding to the power disposal.
[0025] If the second constraint operation meets the first constraint condition, a second objective function is constructed based on the supply amount and quota information given by each of the power supply terminals, wherein the second objective function takes reducing the power disposal cost as the calculation objective.
[0026] Based on the second objective function, the power disposal scheme is constructed.
[0027] As an option, where,
[0028] The second objective function is:
[0029]
[0030] The expression corresponding to the first constraint is:
[0031]
[0032] Where t is time, Indicates the first Whether the request quota from the individual power supplier has been processed, and if so, ,otherwise , The request limit for the power supplier with the highest request limit among all power suppliers that have completed power processing is taken as the power processing limit. The supply quantity corresponding to the supply end i. The power output limit of the supply terminal i.
[0033] Optionally, the power handling scheme based on the constructed power supply utilizes a deep reinforcement learning algorithm to control the power supply end for power supply under the second constraint, including:
[0034] A third constraint operation is performed on the power regulation corresponding to the power supply terminal;
[0035] If the third constraint operation meets the second constraint condition, a third objective function is constructed based on the subsidy operation performed on the power supply side, wherein the third objective function takes controlling the grid operating cost as the calculation objective;
[0036] Based on the third objective function, the power supply end is controlled using the deep reinforcement learning algorithm.
[0037] Optionally, the subsidy operation performed on the power supply side includes:
[0038] A subsidy function is constructed based on the power output cost of the equipment at the power supply end, energy storage battery subsidy information, and / or the input power cost from the main grid.
[0039] The subsidy operation is performed on the power supply side based on the subsidy function.
[0040] As an option, where,
[0041] The expression corresponding to the second constraint is:
[0042]
[0043] The third objective function is:
[0044]
[0045] in, The power output cost of the equipment at the power supply end. Subsidies for the use of energy storage batteries, The cost of electricity input from the main grid varies depending on the real-time electricity price at different times of the day. The adjustment depth subsidy is provided to the power supply end to compensate for losses at the power supply end.
[0046] Optionally, the method further includes:
[0047] Based on the adjustment range of the state of charge of the energy storage at the power supply end, the subsidy information of the energy storage battery at the power supply end is obtained.
[0048] This application also provides a power information processing system based on distributed energy storage, including:
[0049] The prediction module is configured to predict the power load and power supply during the power handling period through an LSTM network and generate corresponding prediction results, wherein the power supply corresponds to multiple power supply terminals, and the power supply terminals are set up in a distributed manner.
[0050] The determination module is configured to determine the required backup energy storage capacity for each unit time period within the power handling time period based on the prediction results, wherein the backup energy storage capacity is associated with the power load;
[0051] The publishing module is configured to publish energy storage demand characterizing the backup energy storage capacity in order to obtain feedback information based on the published energy storage demand.
[0052] The processing module is configured to construct a corresponding power disposal scheme based on the energy storage demand and the feedback information under the first constraint.
[0053] Based on the constructed power disposal scheme, a deep reinforcement learning algorithm is used to control the power supply end for power supply under the second constraint to implement power disposal.
[0054] Optionally, the processing module is further configured as follows:
[0055] Obtain information for adjusting the power disposal plan;
[0056] Under the third constraint, the power disposal plan is adjusted based on the adjustment information.
[0057] Optionally, the processing module is further configured as follows:
[0058] At least the supply quantity corresponding to the power supply terminal and the processing quantity of the supply quantity are subject to a first constraint operation;
[0059] If the first constraint operation meets the third constraint condition, a first objective function is constructed based on the disposal information and disposal results formed during the power transfer process, wherein the first objective function takes increasing the amount of power disposal as the calculation objective.
[0060] The power disposal scheme is adjusted based on the first objective function.
[0061] Optionally, the processing module is further configured as follows:
[0062] A second constraint operation is performed on the power supply quantity corresponding to the power supply end and the power disposal quota corresponding to the power disposal.
[0063] If the second constraint operation meets the first constraint condition, a second objective function is constructed based on the supply amount and quota information given by each of the power supply terminals, wherein the second objective function takes reducing the power disposal cost as the calculation objective.
[0064] Based on the second objective function, the power disposal scheme is constructed.
[0065] Optionally, the processing module is further configured as follows:
[0066] A third constraint operation is performed on the power regulation corresponding to the power supply terminal;
[0067] If the third constraint operation meets the second constraint condition, a third objective function is constructed based on the subsidy operation performed on the power supply side, wherein the third objective function takes controlling the grid operating cost as the calculation objective;
[0068] Based on the third objective function, the power supply terminal is controlled; wherein...
[0069] The processing module is further configured to:
[0070] A subsidy function is constructed based on the power output cost of the equipment at the power supply end, energy storage battery subsidy information, and / or the input power cost from the main grid.
[0071] The subsidy operation is performed on the power supply side based on the subsidy function.
[0072] The power information processing method in this application predicts the load and renewable energy for each time period during the subsequent power disposal period before the power disposal time. Based on the load and output of each time period, a dynamic reserve capacity pool is formed, thereby coordinating the generation side and the power consumption side. This improves the residual regulation capability of distributed energy storage on the power supply side, ensuring the safe and stable operation of the system while maximizing the benefits of the microgrid. Attached Figure Description
[0073] Figure 1 This is a flowchart of a power information processing method based on distributed energy storage, according to an embodiment of this application.
[0074] Figure 2 This is a flowchart of one embodiment of the power information processing method according to this application;
[0075] Figure 3 Examples of embodiments of this application Figure 2 A flowchart of one embodiment of step S70;
[0076] Figure 4 Examples of embodiments of this application Figure 1 A flowchart of an embodiment of step S40;
[0077] Figure 5 Examples of embodiments of this application Figure 1 A flowchart of one embodiment of step S50;
[0078] Figure 6 This is a flowchart of another specific embodiment of the power information processing method according to the present application;
[0079] Figure 7 This is a flowchart illustrating the prediction operation based on an LSTM prediction network according to an embodiment of this application.
[0080] Figure 8 This is a flowchart illustrating the decision-making process using a deep reinforcement learning algorithm, as described in an embodiment of this application.
[0081] Figure 9 This is a structural block diagram of a power information processing system based on distributed energy storage, according to an embodiment of this application. Detailed Implementation
[0082] Various embodiments and features of this application are described herein with reference to the accompanying drawings.
[0083] It should be understood that various modifications can be made to the embodiments described herein. Therefore, the above description should not be considered as limiting, but merely as an example of embodiments. Other modifications within the scope and spirit of this application will be apparent to those skilled in the art.
[0084] The accompanying drawings, which are included in and form part of this specification, illustrate embodiments of the present application and, together with the general description of the present application given above and the detailed description of the embodiments given below, serve to explain the principles of the present application.
[0085] These and other features of this application will become apparent from the following description of preferred forms of embodiments given as non-limiting examples, with reference to the accompanying drawings.
[0086] It should also be understood that although this application has been described with reference to some specific examples, those skilled in the art can certainly implement many other equivalent forms of this application, which have the features described in the claims and are therefore all within the scope of protection defined herein.
[0087] The above and other aspects, features and advantages of this application will become more apparent when taken in conjunction with the accompanying drawings and in view of the following detailed description.
[0088] Specific embodiments of this application are described thereafter with reference to the accompanying drawings; however, it should be understood that the claimed embodiments are merely examples of this application, which can be implemented in various ways. Well-known and / or repeated functions and structures are not described in detail to avoid unnecessary or redundant details that could obscure the application. Therefore, the specific structural and functional details claimed herein are not intended to be limiting, but merely serve as the basis and representative basis for the claims to teach those skilled in the art to use this application in a variety of substantially any suitable detailed structures.
[0089] This specification may use the phrases “in one embodiment,” “in another embodiment,” “in yet another embodiment,” or “in other embodiments,” all of which may refer to one or more of the same or different embodiments according to this application.
[0090] This application presents a power information processing method based on distributed energy storage. This method can be applied to microgrids different from the main power grid, and specifically to a power information processing system. The power information processing system can logically construct an energy storage capacity pool. The generation side can transmit power to this energy storage capacity pool, and the consumption side can obtain power from it. The processing system can dynamically manage and control the energy storage capacity pool and implement appropriate power disposal measures, thereby coordinating the generation side and the consumption side. In this embodiment, power disposal can involve handling the power demand relationship between power load and power supply, including power allocation and specific allocation schemes, thereby coordinating the power output actions of the generation side and the power input actions of the consumption side.
[0091] This power information processing method first predicts the power load and power supply during the power disposal period, such as predicting the power load on the generation side and the power supply on the consumption side, and the power supply on the supply side. This prediction operation can be performed before power disposal is carried out, such as before power allocation between the consumption side and the power supply side.
[0092] Then, based on the prediction results, the required standby energy storage capacity for each unit time period within the power handling period is calculated. The standby energy storage capacity can represent the amount of electricity demand to meet the power load, thereby obtaining the accurate power demand (i.e., energy storage demand) for each time period.
[0093] The system can publish the energy storage demand on the network, allowing users, acting as power suppliers, to provide feedback on their available power capacity. The system can then use this demand and user feedback to set up appropriate power allocation schemes, such as specific plans for power distribution between the consumer and power supplier. During implementation, power allocation can be tailored to the specific circumstances of both the consumer and power supplier to meet their respective needs. For example, by utilizing multiple objective functions with different objectives, the system can maximize power allocation based on the purchase price required by the consumer and the selling price required by the power supplier, while minimizing costs associated with power allocation for all parties. This achieves accurate coordination of interests and avoids power waste.
[0094] Below, in conjunction with the appendix Figure 1 and Figure 6 The power information processing method based on distributed energy storage in this application embodiment includes the following steps:
[0095] S10, the power load and power supply during the power handling period are predicted by the LSTM network, and the corresponding prediction results are generated. The power supply corresponds to multiple power supply terminals, which are set up in a distributed manner.
[0096] Long Short-Term Memory (LSTM) networks are a type of temporal recurrent neural network designed to address the long-term dependency problem inherent in general RNNs (Recurrent Neural Networks). In this embodiment, an LSTM network is used to predict power load and power supply within a specific power handling period. This power handling period can be the timeframe for coordinating power load and power supply; specifically, it can be the coordination period between power consumers with power loads and power generators with power supply capabilities, such as the timeframe for power transactions between consumers and generators. The time unit can be "days," "months," etc.
[0097] When forecasting the power load in a microgrid, predictions can be made using an LSTM network based on time information, historical load data from the power supply, and historical ambient temperature. When forecasting the power supply (energy output) in a microgrid, including the generation capacity of distributed photovoltaic (PV) and distributed wind turbines, predictions can be made based on the historical output and / or solar irradiance of distributed PV, and the historical output and wind speed of distributed wind turbines.
[0098] In one specific embodiment, combined with Figure 7 On the one hand, when using LSTM networks to predict the power load on the consumer side, the input data can be... ,in For time, For the historical load on the electricity consumption side, The corresponding historical ambient temperature, The corresponding day type is divided into weekdays and public holidays. The value is 1 for weekdays and 0 for public holidays. For the corresponding seasons, spring, summer, autumn, and winter take values of 1, 2, 3, and 4 respectively. On the other hand, in power supply forecasting, such as the forecasting of renewable energy output, it is divided into distributed photovoltaic (PV) and distributed wind turbines. The input data for PV output forecasting is... ,in, Contributing to the history of distributed photovoltaic power. The light intensity corresponds to the time period; the input data for output prediction of distributed wind turbines is... ,in, Contributing to the history of distributed photovoltaic power. This represents the wind speed for the corresponding time period. Combining the two forecasting operations described above, we can forecast both power load and power supply, generating corresponding forecast results.
[0099] S20, based on the prediction results, determine the required backup energy storage capacity for each unit time period within the power handling time period, wherein the backup energy storage capacity is associated with the power load.
[0100] During the power response period, the required backup energy storage capacity (electricity demand) varies in different time periods, and the power output from the power supply side may also differ in different time periods. Therefore, in this embodiment, based on the prediction results, the required backup energy storage capacity (electricity demand) for each time period within the power response period can be determined. This allows for more accurate coordination between the power consumption side and the power generation side, ensuring power safety while reducing the overall operating cost of the microgrid.
[0101] In one specific embodiment, the required standby energy storage capacity for each unit time period within the power disposal time period can be determined using the following formula: .in, This is the energy storage backup safety factor for each time period, which can be dynamically adjusted according to actual conditions. This refers to backup energy storage capacity. When the output of renewable energy (power supply) exceeds the power load on the consumer side, the system needs to absorb the excess renewable energy. A positive value indicates a power reserve. Energy storage devices in a microgrid (such as stand-alone energy storage devices or energy storage devices on other power supply sides) provide energy storage capacity. When the output of renewable energy is less than the power load on the consumption side, the system needs to increase the output of the power supply side to maintain power balance. A negative value indicates a reduction in reserves, with energy storage devices in the microgrid providing discharge capacity.
[0102] S30, publish the energy storage demand characterizing the backup energy storage capacity, and obtain feedback information based on the published energy storage demand.
[0103] The standby energy storage capacity corresponds to the electricity demand on the consumer side of the microgrid. The system can publish relevant energy storage demands (such as energy that needs to be stored for later use) before power disposal operations, thus disseminating relevant information about the electricity demand. The power supply side (generation side) can then respond based on this energy storage demand, such as determining its supply based on its own status, generating feedback information, and reporting this feedback information to the system. This allows the system to perform appropriate power disposal based on the feedback information from each power supply side and the energy storage demand.
[0104] S40, under the first constraint, construct a corresponding power disposal scheme based on the energy storage demand and the feedback information.
[0105] The first constraint is a constraint on the supply volume corresponding to the power supply end and the power disposal limit corresponding to the power disposal. The constraint objective of the first constraint is to maintain a balance between the requested limit of all power supply ends that have completed power disposal and the power output limit of the power supply end. Operations that satisfy the first constraint will enable the output power to be disposed of in a timely manner and the power demand to be met in a timely manner.
[0106] The power disposal plan is designed to adapt to the actual energy storage demand and feedback information, thereby reducing disposal costs. Different energy storage demands and feedback information correspond to different power disposal plans, enabling real-time adjustments to the plan and accurate coordination between the generation and consumption sides, thus minimizing disposal costs.
[0107] S50, based on the constructed power disposal scheme, using a deep reinforcement learning algorithm, under the second constraint, control the power supply end for power supply to implement power disposal.
[0108] Deep reinforcement learning algorithms (Deep Deterministic Policy Gradient) combine the perceptual capabilities of deep learning with the decision-making abilities of reinforcement learning, enabling direct control based on input information. Deep learning possesses strong perceptual capabilities but lacks sufficient decision-making ability; while reinforcement learning, though capable of decision-making, is ineffective in perceptual problems. Therefore, deep reinforcement learning algorithms combine the two, achieving complementary advantages and providing a solution to the perceptual decision-making problem in complex systems.
[0109] In this embodiment, the processing system can logically construct an energy storage capacity pool. The power generation side can transmit power to this energy storage capacity pool, and the power consumption side can obtain power from this energy storage capacity pool. The processing system can dynamically manage and control the energy storage capacity pool, and implement power disposal accordingly, thereby coordinating the power generation side and the power consumption side.
[0110] After power disposal, deep reinforcement learning algorithms can be used to calculate the data corresponding to the power disposal plan. Control of the power supply side is then implemented with the goal of minimizing the microgrid's operating cost. Controlling the power supply side allows for the control of the power supply volume (generation) and the charging and discharging of the power supply's battery storage tanks, thereby controlling the power in the energy storage tanks. Furthermore, it enables control of energy interaction between the microgrid and the main grid. This achieves coordination between the generation and consumption sides within the microgrid, as well as coordination between the microgrid and the main grid, further ensuring the interests of all parties and preventing power waste.
[0111] In addition, the second constraint can be a constraint on the output power of the power supply and the input power of the power load, ensuring that the output power of the power supply and the input power of the power load are balanced (e.g., constraining both to be the same).
[0112] The power information processing method described in this implementation predicts the load and renewable energy for each time period within the subsequent power disposal time before the actual power disposal time. Based on the load and output of each time period, a dynamic reserve capacity pool is formed, thereby coordinating the generation and consumption sides. This improves the residual regulation capacity of distributed energy storage on the power supply side, ensuring the safe and stable operation of the system while maximizing the benefits of the microgrid.
[0113] In one embodiment of this application, the control of the power supply terminal for power supply is performed before power processing, such as... Figure 2 As shown, the method further includes:
[0114] S60, Obtain adjustment information for the power handling plan.
[0115] In the short period before the commencement of power processing operations, such as a few hours prior, changes may occur on both the generation and consumption sides. For example, the power supply capacity of distributed photovoltaic and wind turbines on the generation side may temporarily change, and the previously reported supply may not be sufficient. In such cases, adjustments to the original power processing plan are necessary. Similarly, the electricity consumption on the consumption side is likely to exceed the previously reported power load to the system. In this situation, adjustments to the original power processing plan are also required, generating corresponding adjustment information so that the system can access this information.
[0116] S70, under the third constraint, the power disposal plan is adjusted based on the adjustment information.
[0117] The third constraint can be a constraint on the supply of electricity on the supply side and the demand on the demand side. The objective of this third constraint is to maintain a balance between the supply and demand, including controlling the difference between the supply and demand within a small range, ideally with the supply and demand being equal. Under this condition, the electricity handling plan can be adjusted based on adjustment information to ensure the respective interests of both the demand and supply sides.
[0118] In one embodiment of this application, under the third constraint condition, the power disposal plan is adjusted based on the adjustment information, such as... Figure 3 As shown, it includes the following steps:
[0119] S710, at least the supply quantity corresponding to the power supply terminal and the processing quantity for processing the supply quantity are subject to a first constraint operation;
[0120] S720, if the first constraint operation meets the third constraint condition, a first objective function is constructed based on the disposal information and disposal results formed during the power flow process, wherein the first objective function takes increasing the amount of power disposal as the calculation objective;
[0121] S730, the power disposal scheme is adjusted based on the first objective function.
[0122] Specifically, the first constraint operation includes operations that constrain the supply volume corresponding to the power supply side and the demand volume corresponding to the power consumption side, as well as operations that constrain the disposal volume of the supply volume. For example, the supply volume and the output amount of the supply volume (including the economic benefit amount) can form the first data, and the demand volume and the input amount of the demand volume (including the economic benefit amount) can form the second data. The first constraint operation can constrain the first data and the second data so that the respective demands (including economic demands) of the power consumption side and the power generation side are met.
[0123] Accordingly, the third constraint includes constraints on the supply volume corresponding to the power supply side and the demand volume corresponding to the power consumption side, as well as constraints on the disposal volume of the supply volume. Given that the first constraint operation meets the third constraint, a first objective function can be constructed based on the disposal information generated during the power flow process (such as the input quota on the power consumption side and the output quota on the power generation side) and the disposal results (such as whether the power consumption side and the power generation side have achieved power allocation). The first objective function aims to increase the amount of power disposed of. Under the third constraint, as much power disposal as possible can be achieved. Therefore, after adjusting the power disposal scheme based on the first objective function, the adjusted power disposal scheme can be more suitable for the current state of the power generation side and the power consumption side.
[0124] In one embodiment of this application, wherein,
[0125] The first objective function is:
[0126]
[0127] The expression corresponding to the third constraint condition is:
[0128]
[0129] Where t is the time parameter, and These are the input and output quotas generated during the power disposal process. The input quota can be the amount of electricity that the power user can draw from the energy storage pool, such as the quantity or price of electricity requested from the pool. The output quota, on the other hand, can be the amount of electricity that the power generator can inject into the energy storage pool, such as the quantity or price of electricity injected or sold.
[0130] and These represent whether power-related actions based on the introduced and output quotas have been implemented, respectively. A value of 1 indicates implementation, while a value of 0 indicates non-implementation. For the electricity disposal quota, The supply quantity corresponding to the supply end i. The processing volume corresponding to the supply end j can be a quantified value for achieving energy flow. This indicates a constraint operation.
[0131] In one embodiment of this application, under the first constraint, a corresponding power disposal scheme is constructed based on the energy storage demand and the feedback information, such as... Figure 4 As shown, it includes the following steps:
[0132] S410, perform a second constraint operation on the power supply quantity corresponding to the power supply end and the power disposal quota corresponding to the power disposal;
[0133] S420, if the second constraint operation meets the first constraint condition, a second objective function is constructed based on the supply amount and quota information given by each of the power supply terminals, wherein the second objective function takes reducing the power disposal cost as the calculation objective;
[0134] S430, Based on the second objective function, construct the power disposal scheme.
[0135] Specifically, the second constraint operation is a constraint operation on the supply and the corresponding power disposal quota. The constraint objective of the second constraint operation is to maintain a balance between the requested quota of all power suppliers that have completed power disposal and the power output quota of the power suppliers, so that the output power can be disposed of in a timely manner and the power demand can be met in a timely manner.
[0136] Accordingly, the first constraint condition is a condition that constrains the supply volume corresponding to the power supply end and the power disposal quota corresponding to the power disposal. For example, the first constraint condition could be a condition that constrains the supply volume of power input into the energy storage capacity pool from the generation side and the quantity or clearing amount of the power disposed of. If it is determined that the second constraint operation meets the first constraint condition, a second objective function can be constructed based on the supply volume and quota information provided by each power supply end, wherein the second objective function takes reducing the power disposal cost as the calculation objective.
[0137] In one embodiment of this application, wherein,
[0138] The second objective function is:
[0139]
[0140] The expression corresponding to the first constraint is:
[0141]
[0142] Where t is time, Indicates the first Whether the request quota from the individual power supplier has been processed, and if so, ,otherwise The requested amount can be a quantity of electricity-related items requested by the power supplier, such as the quantity of electricity to be allocated or the electricity price.
[0143] The clearing price is determined by the highest requested amount among all power suppliers that have completed power disposal. For example, among all power suppliers that have sold electricity (generation side), the highest bid is used as the clearing price. .
[0144] The supply quantity corresponding to the supply end i. This refers to the power output limit of the power supply terminal i. The output limit can be the specific value of the amount of electricity that the power supply terminal can output, that is, the energy storage capacity that can be provided.
[0145] In one embodiment of this application, such as Figure 5 As shown, the power management scheme based on the constructed power supply utilizes a deep reinforcement learning algorithm to control the power supply end for power supply under a second constraint, including:
[0146] S510, perform a third constraint operation on the power regulation corresponding to the power supply terminal;
[0147] S520, if the third constraint operation meets the second constraint condition, a third objective function is constructed based on the subsidy operation performed on the power supply side, wherein the third objective function takes controlling the grid operating cost as the calculation objective;
[0148] S530, based on the third objective function, the power supply end is controlled using the deep reinforcement learning algorithm.
[0149] Specifically, the third constraint operation can be an operation that constrains the output power of the power supply and the input power of the power load, ensuring that the difference between the output power of the power supply and the input power of the power load falls within a preset range, or that they are equal. The third constraint operation achieves power balance between the generation and consumption sides, preventing safety accidents and ensuring that the electricity generated by the generation side is not wasted.
[0150] Correspondingly, the second constraint can be a constraint on the output power of the power supply and the input power of the power load, ensuring a balance between the output power of the power supply and the input power of the power load (e.g., constraining both to be the same). If the third constraint operation is determined to meet the second constraint, a third objective function is constructed based on the subsidy operation performed on the power supply, wherein the third objective function aims to control the grid operating cost. This ensures that, in the process of controlling the power supply based on the third objective function using a deep reinforcement learning algorithm, both supply and demand balance can be guaranteed while reducing the operating cost of the microgrid.
[0151] Furthermore, in one embodiment, combined with Figure 8 The steps for making decisions using the Deep Reinforcement Learning (DDPG) algorithm include:
[0152] First, construct the current value Actor and Critic network and the target value Actor and Critic network, and initialize their weights and biases, and randomly initialize the current state.
[0153] Second, initialize the random process used for training, and based on the observed states, obtain the actions in the current environment using the current state and the current value of the Actor network. Among them, random perturbations are added when generating actions.
[0154] Third, execute the action and obtain the next state and reward. .
[0155] Fourth, store the state transitions, actions, and rewards in the above process into the experience replay pool.
[0156] Fifth, random sampling from the experience replay pool. For each sample, the Critic network is updated with the goal of minimizing the loss function, and the Actor network is updated with the current value using policy gradient descent. The target Critic network and Actor network are updated using the parameters of the current value network.
[0157] Sixth, when the network loss function stabilizes or reaches the set number of training steps, the training is completed, and the trained network is used to perform optimized control of power disposal operations.
[0158] In one embodiment of this application, the subsidy operation performed on the power supply side includes the following steps:
[0159] A subsidy function is constructed based on the power output cost of the equipment at the power supply end, energy storage battery subsidy information, and / or the input power cost from the main grid.
[0160] The subsidy operation is performed on the power supply side based on the subsidy function.
[0161] Specifically, on the one hand, power supply equipment incurs losses during power supply, especially older equipment, which experiences greater losses per unit of power supply time. On the other hand, the losses from the same equipment may vary depending on the region. Based on this, the method in this application provides subsidies to the power supply side to compensate for the loss due to the shortened lifespan of energy storage on the generation side. Specifically, this includes constructing a subsidy function based on the power supply side's equipment output cost, energy storage battery subsidy information, and / or the input power cost from the main grid. The energy storage battery subsidy information refers to information on subsidies for energy storage batteries (charging and discharging) used by the power supply side; the input power cost is the cost of obtaining electricity from the main grid (national or regional backbone grid); and the corresponding subsidy function is:
[0162]
[0163] in, To provide deep subsidies for regulating the power supply side, Adjust the depth subsidy coefficient, For the power supply side during the regulation process The maximum state of charge of the stored energy. For the power supply side during the regulation process The minimum state of charge of the energy storage. Once this subsidy function is obtained, subsidies can be applied to the power supply side based on this function.
[0164] In one embodiment of this application, the method further includes: obtaining energy storage battery subsidy information of the power supply terminal based on the adjustment range of the state of charge of the energy storage at the power supply terminal.
[0165] This adjustment range can refer to the range within which the power supply adjusts the energy storage capacity. The power supply stores generated electricity to provide power to the consumer side based on power consumption. This adjustment range can also be the range between the minimum and maximum amount of energy storage. The energy storage battery subsidy information for the power supply can be constructed based on this range, representing the relevant information requiring subsidies for the power supply. Subsidy operations can then be performed on the power supply based on this energy storage battery subsidy information.
[0166] Furthermore, in one specific embodiment of this application, the expression corresponding to the second constraint condition is:
[0167]
[0168] The third objective function is:
[0169]
[0170] in, The power output cost of the equipment at the power supply end, i.e., the cost of generating electricity. Subsidies for the use of energy storage batteries, The cost of electricity input from the main grid varies depending on the real-time electricity price at different times of the day. The adjustment depth subsidy is provided to the power supply end to compensate for losses at the power supply end.
[0171] This application also provides a power information processing system based on distributed energy storage, such as... Figure 9 As shown, it includes:
[0172] The prediction module is configured to predict the power load and power supply during the power handling period through an LSTM network and generate corresponding prediction results, wherein the power supply corresponds to multiple power supply terminals, which are distributed.
[0173] Long Short-Term Memory (LSTM) networks are a type of temporal recurrent neural network designed to address the long-term dependency problem inherent in general RNNs (Recurrent Neural Networks). In this embodiment, the prediction module uses an LSTM network to predict power load and power supply within a specific power handling period. This power handling period can be a timeframe for coordinating power load and power supply; specifically, it can be a timeframe for coordination between power consumers with power loads and power generators with power supply capabilities, such as the timeframe for power transactions between consumers and generators. The time unit can be "days," "months," etc.
[0174] When forecasting the power load in a microgrid, the forecasting module can use an LSTM network to make predictions based on time information, historical load data from the power supply, and historical ambient temperature. When forecasting the power supply (energy output) in the microgrid, including the generation capacity of distributed photovoltaic (PV) and distributed wind turbines, predictions can be made based on the historical output and / or solar irradiance of distributed PV, and the historical output and wind speed of distributed wind turbines.
[0175] In one specific embodiment, when the prediction module uses an LSTM network to predict the power load on the power consumption side, the input data can be... ,in For time, For the historical load on the electricity consumption side, The corresponding historical ambient temperature, The corresponding day type is divided into weekdays and public holidays. The value is 1 for weekdays and 0 for public holidays. For the corresponding seasons, spring, summer, autumn, and winter take values of 1, 2, 3, and 4 respectively. On the other hand, the forecasting module predicts electricity supply, such as the output of renewable energy sources, which is divided into distributed photovoltaic (PV) and distributed wind turbines. The input data for PV output forecasting is... ,in, Contributing to the history of distributed photovoltaic power. The light intensity corresponds to the time period; the input data for output prediction of distributed wind turbines is... ,in, Contributing to the history of distributed photovoltaic power. This represents the wind speed for the corresponding time period. Combining the two forecasting operations described above, we can forecast both power load and power supply, generating corresponding forecast results.
[0176] The determination module is configured to determine the required backup energy storage capacity for each unit time period within the power handling time period based on the prediction results, wherein the backup energy storage capacity is associated with the power load.
[0177] During the power response period, the required backup energy storage capacity (electricity demand) varies in different time periods, and the power output from the power supply side may also differ in different time periods. Therefore, in this embodiment, the determination module determines the required backup energy storage capacity (electricity demand) for each time period within the power response period based on the prediction results. This allows for more accurate coordination between the power consumption side and the power generation side, ensuring power safety while reducing the overall operating cost of the microgrid.
[0178] In one specific embodiment, the determining module can use the following formula to determine the required standby energy storage capacity for each unit time period within the power handling time period: .in, This is the energy storage backup safety factor for each time period, which can be dynamically adjusted according to actual conditions. This refers to backup energy storage capacity. When the output of renewable energy (power supply) exceeds the power load on the consumer side, the system needs to absorb the excess renewable energy. A positive value indicates a power reserve. Energy storage devices in a microgrid (such as stand-alone energy storage devices or energy storage devices on other power supply sides) provide energy storage capacity. When the output of renewable energy is less than the power load on the consumption side, the system needs to increase the output of the power supply side to maintain power balance. A negative value indicates a reduction in reserves, with energy storage devices in the microgrid providing discharge capacity.
[0179] The publishing module is configured to publish energy storage demand that represents the backup energy storage capacity, in order to obtain feedback information based on the published energy storage demand.
[0180] The standby energy storage capacity represents the electricity demand on the consumer side of the microgrid. The publishing module can publish the corresponding energy storage demand (such as energy that needs to be stored for later use) before power disposal operations, thus disseminating relevant information about the electricity demand. The power supply side (generation side) can then respond based on this energy storage demand, such as determining its supply based on its own status, generating feedback information, and reporting this feedback information to the system. This allows the system to perform appropriate power disposal based on the feedback information from each power supply side and the energy storage demand.
[0181] The processing module is configured to construct a corresponding power disposal scheme based on the energy storage demand and the feedback information under the first constraint.
[0182] Based on the constructed power disposal scheme, a deep reinforcement learning algorithm is used to control the power supply end for power supply under the second constraint to implement power disposal.
[0183] Specifically, the first constraint is a constraint on the supply volume corresponding to the power supply end and the power disposal quota corresponding to the power disposal. The constraint objective of the first constraint is to maintain a balance between the requested quota of all power supply ends that have completed power disposal and the power output quota of the power supply end. Operations that satisfy the first constraint will enable the output power to be disposed of in a timely manner and the power demand to be met in a timely manner.
[0184] The power disposal plan is designed to adapt to the actual energy storage demand and feedback information, thereby reducing disposal costs. Different energy storage demands and feedback information correspond to different power disposal plans, enabling real-time adjustments to the plan and accurate coordination between the generation and consumption sides, thus minimizing disposal costs.
[0185] Deep reinforcement learning (Deep Deterministic Policy Gradient) algorithms combine the perceptual capabilities of deep learning with the decision-making abilities of reinforcement learning, enabling direct control based on the input image. Deep learning possesses strong perceptual capabilities but lacks sufficient decision-making ability; while reinforcement learning, though capable of decision-making, is ineffective in perceptual problems. Therefore, deep reinforcement learning algorithms combine the two, achieving complementary advantages and providing a solution to the perceptual decision-making problem in complex systems.
[0186] In this embodiment, the system can logically construct an energy storage capacity pool. The power generation side can transmit power to this energy storage capacity pool, and the power consumption side can obtain power from this energy storage capacity pool. The processing module can dynamically manage and control the energy storage capacity pool, and implement power disposal accordingly, thereby coordinating the power generation side and the power consumption side.
[0187] After power disposal, the processing module can utilize deep reinforcement learning algorithms to calculate the data corresponding to the power disposal plan. It controls the power supply side with the goal of minimizing microgrid operating costs. Controlling the power supply side allows for control over the power supply volume (generation) and the charging and discharging of the power supply's battery storage tanks, thereby controlling the power in the energy storage tanks. Furthermore, it can control the energy interaction between the microgrid and the main grid. This achieves coordination between the generation and consumption sides within the microgrid, as well as coordination between the microgrid and the main grid, further ensuring the interests of all parties and preventing power waste.
[0188] In addition, the second constraint can be a constraint on the output power of the power supply and the input power of the power load, ensuring that the output power of the power supply and the input power of the power load are balanced (e.g., constraining both to be the same).
[0189] In one embodiment of this application, the processing module is further configured to:
[0190] Obtain information for adjusting the power disposal plan;
[0191] Under the third constraint, the power disposal plan is adjusted based on the adjustment information.
[0192] Specifically, in the short period before the commencement of power processing operations, such as a few hours prior, changes may occur on both the generation and consumption sides. For instance, the power supply capacity of distributed photovoltaic and wind turbines on the generation side may temporarily change, and the previously reported supply may not be sufficient. In this case, the processing module needs to adjust the original power handling plan. Similarly, the electricity consumption on the consumption side is likely to exceed the previously reported power load. In this situation, the processing module also needs to adjust the original power handling plan and generate corresponding adjustment information so that the system can access this information.
[0193] The third constraint can be a constraint on the supply volume of electricity on the supply side and the demand volume on the demand side. The objective of this third constraint is to maintain a balance between the supply volume and the demand volume, including controlling the difference between the supply volume and the demand volume within a small range, ideally with the supply volume and demand volume being the same. Under this condition, the processing module adjusts the power disposal plan based on the adjustment information to ensure the respective interests of the power supply and demand sides.
[0194] In one embodiment of this application, the processing module is further configured to:
[0195] At least the supply quantity corresponding to the power supply terminal and the processing quantity of the supply quantity are subject to a first constraint operation;
[0196] If the first constraint operation meets the third constraint condition, a first objective function is constructed based on the disposal information and disposal results in the power transfer process, wherein the first objective function takes increasing the amount of power disposal as the calculation objective;
[0197] The power disposal scheme is adjusted based on the first objective function.
[0198] Specifically, the first constraint operation includes operations that constrain the supply volume corresponding to the power supply side and the demand volume corresponding to the power consumption side, as well as operations that constrain the disposal volume of the supply volume. For example, the supply volume and the output amount of that supply volume (including the economic benefit amount) can form the first data, and the demand volume and the input amount of that demand volume (including the economic benefit amount) can form the second data. The processing module can constrain the first data and the second data through the first constraint operation so that the respective demands (including economic demands) of both the power consumption side and the power generation side are met.
[0199] Accordingly, the third constraint includes constraints on the supply volume corresponding to the power supply side and the demand volume corresponding to the power consumption side, as well as constraints on the disposal volume of the supply volume. When the processing module determines that the first constraint operation meets the third constraint, it can construct a first objective function based on the disposal information during the power flow process (such as the input quota on the power consumption side and the output quota on the power generation side) and the disposal results (such as whether power allocation has been achieved between the power consumption side and the power generation side). The first objective function aims to increase the amount of power disposed of. Under the third constraint, as much power disposal as possible can be achieved. Therefore, after adjusting the power disposal scheme based on the first objective function, the adjusted power disposal scheme can be more suitable for the current state of the power generation side and the power consumption side.
[0200] In one embodiment of this application, the processing module is further configured to:
[0201] A second constraint operation is performed on the power supply quantity corresponding to the power supply end and the power disposal quota corresponding to the power disposal.
[0202] If the second constraint operation meets the first constraint condition, a second objective function is constructed based on the supply amount and quota information given by each of the power supply terminals, wherein the second objective function takes reducing the power disposal cost as the calculation objective.
[0203] Based on the second objective function, the power disposal scheme is constructed.
[0204] Specifically, the second constraint operation is a constraint operation on the supply and the corresponding power disposal quota. The constraint objective of the second constraint operation is to maintain a balance between the requested quota of all power suppliers that have completed power disposal and the power output quota of the power suppliers, so that the output power can be disposed of in a timely manner and the power demand can be met in a timely manner.
[0205] Accordingly, the first constraint condition is a condition that constrains the supply volume corresponding to the power supply end and the power disposal quota corresponding to the power disposal. For example, the first constraint condition could be a condition that constrains the supply volume of power input into the energy storage capacity pool from the generation side and the quantity or clearing amount of the power disposed of. If the processing module determines that the second constraint operation meets the first constraint condition, it can then construct a second objective function based on the supply volume and quota information provided by each power supply end, wherein the second objective function aims to reduce the power disposal cost.
[0206] In one embodiment of this application, the processing module is further configured to:
[0207] A third constraint operation is performed on the power regulation corresponding to the power supply terminal;
[0208] If the third constraint operation meets the second constraint condition, a third objective function is constructed based on the subsidy operation performed on the power supply side, wherein the third objective function takes controlling the grid operating cost as the calculation objective;
[0209] Based on the third objective function, the power supply terminal is controlled; wherein...
[0210] The processing module is further configured to:
[0211] A subsidy function is constructed based on the power output cost of the equipment at the power supply end, energy storage battery subsidy information, and / or the input power cost from the main grid.
[0212] The subsidy operation is performed on the power supply side based on the subsidy function.
[0213] Specifically, the third constraint operation can be an operation that constrains the output power of the power supply and the input power of the power load, ensuring that the difference between the output power of the power supply and the input power of the power load falls within a preset range, or that they are equal. The third constraint operation achieves power balance between the generation and consumption sides, preventing safety accidents and ensuring that the electricity generated by the generation side is not wasted.
[0214] Correspondingly, the second constraint can be a constraint on the output power of the power supply and the input power of the power load, ensuring a balance between the output power of the power supply and the input power of the power load (e.g., constraining both to be the same). If the processing module determines that the third constraint operation meets the second constraint, it constructs a third objective function based on the subsidy operation performed on the power supply, wherein the third objective function aims to control the grid operating cost. This allows the processing module to maintain supply and demand balance while reducing the microgrid's operating cost during the process of controlling the power supply based on the third objective function and utilizing a deep reinforcement learning algorithm.
[0215] Furthermore, on the one hand, power supply equipment incurs losses during power supply, especially older equipment, which experiences greater losses per unit of power supply time. On the other hand, the losses from the same equipment may vary depending on the region. Based on this situation, the processing module of this application performs a subsidy operation on the power supply side that provides power, thereby compensating for the loss due to the reduced lifespan of energy storage on the generation side. Specifically, the processing module constructs a subsidy function based on the power supply side's equipment output cost, energy storage battery subsidy information, and / or the input power cost from the main grid. Here, the energy storage battery subsidy information refers to information on subsidies for using energy storage batteries (charging and discharging) from the power supply side, the input power cost is the cost of obtaining power from the main grid (national or regional backbone grid), and the corresponding subsidy function is:
[0216]
[0217] in, To provide deep subsidies for regulating the power supply side, Adjust the depth subsidy coefficient, For the power supply side during the regulation process The maximum state of charge of the stored energy. For the power supply side during the regulation process The minimum state of charge of the energy storage. Once this subsidy function is obtained, subsidies can be applied to the power supply side based on this function.
[0218] The above embodiments are merely exemplary embodiments of this application and are not intended to limit this application. The scope of protection of this application is defined by the claims. Those skilled in the art can make various modifications or equivalent substitutions to this application within its substance and scope of protection, and such modifications or equivalent substitutions should also be considered to fall within the scope of protection of this application.
Claims
1. A power information processing method based on distributed energy storage, characterized in that, include: The power load and power supply during the power handling period are predicted by the LSTM network, and the corresponding prediction results are generated. The power supply corresponds to multiple power supply terminals, which are set up in a distributed manner. Based on the prediction results, the required backup energy storage capacity for each unit time period within the power handling time period is determined, wherein the backup energy storage capacity is associated with the power load; Publish energy storage demand characterizing the backup energy storage capacity, and obtain feedback information based on the published energy storage demand; Under the first constraint, a corresponding power disposal scheme is constructed based on the energy storage demand and the feedback information, wherein the first constraint is a condition that constrains the supply volume corresponding to the power supply end and the power disposal quota corresponding to the power disposal. Based on the constructed power disposal scheme, a deep reinforcement learning algorithm is used to control the power supply terminal under a second constraint to implement power disposal. This second constraint is a condition that constrains the output power of the power supply terminal and the input power of the power load. The method further includes, prior to controlling the power supply terminal for power supply to implement power disposal: Obtain information for adjusting the power disposal plan; Under the third constraint, the power disposal plan is adjusted based on the adjustment information, wherein the third constraint includes a constraint on the supply quantity corresponding to the power supply side and the demand quantity corresponding to the power consumption side, and a constraint on the disposal quantity of the supply quantity. The adjustment of the power disposal plan based on the adjustment information under the third constraint includes: At least the supply quantity corresponding to the power supply terminal and the processing quantity of the supply quantity are subject to a first constraint operation; If the first constraint operation meets the third constraint condition, a first objective function is constructed based on the disposal information and disposal results formed during the power transfer process, wherein the first objective function is a function that aims to increase the amount of power disposal. The power disposal plan is adjusted based on the first objective function; Under the first constraint, constructing a corresponding power disposal scheme based on the energy storage demand and the feedback information includes: A second constraint operation is performed on the power supply quantity corresponding to the power supply end and the power disposal quota corresponding to the power disposal. If the second constraint operation meets the first constraint condition, a second objective function is constructed based on the supply quantity and quota information given by each of the power supply terminals, wherein the second objective function is a function with the calculation objective of reducing the power disposal cost; Based on the second objective function, the power disposal scheme is constructed; The constructed power management scheme, utilizing a deep reinforcement learning algorithm, controls the power supply end for power supply under a second constraint, including: A third constraint operation is performed on the power regulation corresponding to the power supply terminal; If the third constraint operation meets the second constraint condition, a third objective function is constructed based on the subsidy operation performed on the power supply side, wherein the third objective function is a function with the calculation objective of controlling the grid operating cost; Based on the third objective function, the deep reinforcement learning algorithm is used to control the power supply end; The subsidy operation performed on the power supply side includes: A subsidy function is constructed based on the equipment output cost of the power supply end, energy storage battery subsidy information, and / or the input power cost from the main grid, wherein the subsidy function is a function used to perform subsidy operations on the power supply end; The subsidy operation is performed on the power supply side based on the subsidy function.
2. The method according to claim 1, characterized in that, in, The first objective function is: The expression corresponding to the third constraint condition is: Where t is time, and These refer to the import and export quotas generated during the power disposal process. and These represent whether power-related actions based on the introduced and output quotas have been implemented; a value of 1 indicates implementation, and a value of 0 indicates non-implementation. For the electricity disposal quota, The supply quantity corresponding to the supply end i. The disposal quantity corresponding to supply end j. This indicates a constraint operation.
3. The method according to claim 1, characterized in that, in, The second objective function is: The expression corresponding to the first constraint is: Where t is time, Indicates the first Whether the request quota from the individual power supplier has been processed, and if so, ,otherwise , The request limit for the power supplier with the highest request limit among all power suppliers that have completed power processing is taken as the power processing limit. The supply quantity corresponding to the supply end i. The power output limit of the supply terminal i.
4. The method according to claim 1, characterized in that, in, The expression corresponding to the second constraint is: The third objective function is: in, The power output cost of the equipment at the power supply end. Subsidies for the use of energy storage batteries, The cost of electricity input from the main grid varies depending on the real-time electricity price at different times of the day. The adjustment depth subsidy is provided to the power supply end to compensate for losses at the power supply end.
5. The method according to claim 1, characterized in that, The method further includes: Based on the adjustment range of the state of charge of the energy storage at the power supply end, the subsidy information of the energy storage battery at the power supply end is obtained.
6. A power information processing system based on distributed energy storage, characterized in that, include: The prediction module is configured to predict the power load and power supply during the power handling period through an LSTM network and generate corresponding prediction results, wherein the power supply corresponds to multiple power supply terminals, and the power supply terminals are set up in a distributed manner. The determination module is configured to determine the required backup energy storage capacity for each unit time period within the power handling time period based on the prediction results, wherein the backup energy storage capacity is associated with the power load; The publishing module is configured to publish energy storage demand characterizing the backup energy storage capacity in order to obtain feedback information based on the published energy storage demand. The processing module is configured to construct a corresponding power disposal scheme based on the energy storage demand and the feedback information under a first constraint condition, wherein the first constraint condition is a condition that constrains the power supply quantity corresponding to the power supply end and the power disposal quota corresponding to the power disposal. Based on the constructed power disposal scheme, a deep reinforcement learning algorithm is used to control the power supply terminal under a second constraint to implement power disposal. This second constraint is a condition that constrains the output power of the power supply terminal and the input power of the power load. The processing module is further configured as follows: Obtain information for adjusting the power disposal plan; Under the third constraint, the power disposal plan is adjusted based on the adjustment information, wherein the third constraint includes a constraint on the supply quantity corresponding to the power supply side and the demand quantity corresponding to the power consumption side, and a constraint on the disposal quantity of the supply quantity. The processing module is further configured as follows: At least the supply quantity corresponding to the power supply terminal and the processing quantity of the supply quantity are subject to a first constraint operation; If the first constraint operation meets the third constraint condition, a first objective function is constructed based on the disposal information and disposal results formed during the power transfer process, wherein the first objective function is a function that aims to increase the amount of power disposal. The power disposal plan is adjusted based on the first objective function; The processing module is further configured as follows: A second constraint operation is performed on the power supply quantity corresponding to the power supply end and the power disposal quota corresponding to the power disposal. If the second constraint operation meets the first constraint condition, a second objective function is constructed based on the supply quantity and quota information given by each of the power supply terminals, wherein the second objective function is a function with the calculation objective of reducing the power disposal cost; Based on the second objective function, the power disposal scheme is constructed; The processing module is further configured as follows: A third constraint operation is performed on the power regulation corresponding to the power supply terminal; If the third constraint operation meets the second constraint condition, a third objective function is constructed based on the subsidy operation performed on the power supply side, wherein the third objective function is a function with the calculation objective of controlling the grid operating cost; Based on the third objective function, the power supply terminal is controlled; wherein... The processing module is further configured to: A subsidy function is constructed based on the equipment output cost of the power supply end, energy storage battery subsidy information, and / or the input power cost from the main grid, wherein the subsidy function is a function used to perform subsidy operations on the power supply end; The subsidy operation is performed on the power supply side based on the subsidy function.