Intelligent micro-grid low-carbon energy scheduling method and system under security constraint
By introducing a secure network and carbon trading mechanism into the smart microgrid, the combined electricity-carbon goal is established, and the problems of insufficient adaptation to the randomness of renewable energy and neglecting safety constraints in the existing technology are solved, and the security of smart microgrids is achieved. Carbon energy scheduling.
Patent Information
- Application Number
- CN202410928480.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-11
- Publication Date
- 2024-11-15
AI Technical Summary
The existing technology is difficult to effectively adapt to the high randomness of renewable energy in smart microgrids, resulting in the optimization results being too conservative, easily falling into local optimal solutions, and ignores the safety operation constraints of the power system, and most methods do not consider carbon emissions. question.
Introduce a secure network, guide the strategy network learning to meet the security operation constraints during offline training, build an electric-carbon joint goal through the carbon trading mechanism, convert low-carbon optimization scheduling problems into Markov decision-making process, and use safe energy scheduling agents to achieve Online low-carbon energy scheduling.
It realizes the safe and low-carbon operation of the smart microgrid, meets the safe operation constraints of energy units, and incentivizes the system to operate in a low-carbon mode, avoids the increase in carbon emissions, and ensures the safety of energy storage equipment and main grid transactions.
Smart Images

Figure CN118970885A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of power system energy optimization and dispatching, and in particular to a smart microgrid low-carbon energy dispatching method and system under safety constraints. Background Art
[0002] In recent years, the problem of global warming caused by excessive carbon dioxide emissions has become an urgent problem for human society. Smart microgrids that integrate a large amount of renewable energy can effectively promote the development of power systems towards green and clean methods. However, renewable energy and load demand have great dynamic randomness, which brings great challenges to the energy management of microgrids. Therefore, low-carbon energy scheduling methods that can dynamically adapt to various random factors of smart microgrids have become a research hotspot.
[0003] At present, the technologies for solving microgrid scheduling include linear programming methods, robust optimization algorithms, heuristic algorithms, etc. However, the above methods cannot adapt well to the high randomness of renewable energy, which will lead to problems such as overly conservative optimization results, easy to fall into local optimal solutions, and difficult to expand. With the continuous advancement of artificial intelligence technology, deep reinforcement learning algorithms have become a hot topic in the field of microgrid energy scheduling. It regards the energy scheduling problem of microgrids as a sequence of decision-making problems, learns the optimal control strategy through the continuous interaction between the intelligent agent and the microgrid environment, can adapt well to uncertain factors, and has good scalability.
[0004] Some existing patents, such as the patent with the authorization announcement number CN113139682B and the invention name "A microgrid energy management method based on deep reinforcement learning", propose a microgrid energy management and optimization method based on Dueling DQN deep reinforcement learning. This method outputs discrete control actions and ignores the operation constraints of the microgrid. There are many safe operation constraints in the power system. The violation of the constraints will bring operational risks to the microgrid system. Therefore, it is extremely important to reasonably dispatch the output of energy units and ensure the safe operation constraints. The patent with the application publication number CN116247648A and the invention name "A deep reinforcement learning method for microgrid energy scheduling under source and load uncertainty" proposes an energy scheduling method based on a deep deterministic policy gradient algorithm. It adds the constraint violation of the energy storage device as a penalty term to the reward function to guide the strategy to ensure the safety constraints of the energy storage device, but this method relies on the fine modeling of the reward term and the penalty term and is not universal. The patent application publication number is CN116995645A, and the invention name is "Power System Security Constraint Economic Dispatch Method Based on Protection Mechanism Reinforcement Learning". It adds a safety constraint layer at the end of the proximal strategy optimization strategy network to avoid dangerous actions. The correction of dangerous actions by this method is carried out in the environment and cannot be continuously updated during the training process of the intelligent agent, which may lead to inaccurate behavior correction. In addition, most microgrid dispatching methods based on deep reinforcement learning mainly consider the economic benefits of microgrids, without considering carbon emissions.
[0005] In summary, it is necessary to design a low-carbon energy scheduling method that meets the safety constraints of microgrid operation. Summary of the invention
[0006] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a low-carbon energy scheduling method and system for a smart microgrid under safety constraints, introduce a safety network, and guide the strategy network to learn in the direction of satisfying safety operation constraints during offline training, so as to achieve safe and low-carbon operation of the smart microgrid.
[0007] In a first aspect, a method for low-carbon energy dispatching of a smart microgrid under security constraints is provided, comprising the following steps:
[0008] S1: Set safe operation constraints according to the microgrid structure and energy unit operating conditions;
[0009] S2: Calculate the cost according to the operation status of the microgrid and set the basic optimization target;
[0010] S3: Introduce carbon trading into microgrid operations and establish a joint electricity-carbon target;
[0011] S4: Convert the low-carbon optimization scheduling problem into a Markov decision process;
[0012] S5: Introduce a safety network to build a safe energy dispatching agent, and perform offline training of the safe energy dispatching agent based on the historical operation data of the microgrid and the Markov decision process;
[0013] S6: Use the safe energy dispatching agent to interact with the microgrid environment in real time to achieve online low-carbon energy dispatching of the microgrid.
[0014] Furthermore, in step S1, the safe operation constraints include at least one of the power constraints of the energy unit, the energy supply and demand balance constraints, the power transaction constraints with the main power grid and the charge state constraints of the energy storage device.
[0015] Furthermore, in step S2, the basic optimization target includes the energy cost generated by the transaction between the microgrid and the main grid, the energy cost generated by the transaction between the microgrid and the natural gas grid, and the investment and operation cost of the energy unit.
[0016] Furthermore, in step S3, a carbon trading mechanism is introduced to quantify the carbon trading cost, which is specifically expressed as:
[0017]
[0018] In the formula, C carbon,t represents the carbon trading cost, represents the price of carbon trading, Indicates the amount of electricity purchased from the main power grid. Indicates the output power of the gas boiler. and They represent the unit power carbon emission factor and carbon quota coefficient of electricity purchased from the main grid, and They respectively represent the unit power carbon emission factor and carbon quota coefficient of the gas boiler.
[0019] Further, in step S4, the state is represented as Actions are represented by The reward function is expressed as r(s t ,a t )=-(C grid,t +C gas,t +C invest,t +C carbon,t ), where t represents the time, and They represent the price of selling electricity to the main grid, the price of purchasing electricity from the main grid, the price of purchasing natural gas, and the price of carbon trading. and They represent the heat energy demand and electricity energy demand of the user at time t, represents the power generation of the photovoltaic power generation equipment at time t, and They represent the charge states of the thermal energy storage device and the electrical energy storage device at time t, respectively. Indicates the charging and discharging power of the electrical energy storage device. and Represent the thermal power of thermal energy storage equipment, electric boiler and gas boiler respectively, C grid,t represents the energy cost generated by the microgrid and the main grid, C gas,t represents the energy cost generated by the transaction between the microgrid and the natural gas grid, C invest,t represents the investment and operation cost of the energy unit, C carbon,t represents the carbon trading cost;
[0020] The safety cost function is introduced to represent the satisfaction of the safety operation constraints. The safety cost function includes the following safety cost functions: At least one of:
[0021]
[0022]
[0023] Where, [x] + =max(0,x), indicates that the safety operation constraints described are met, Indicates the amount of electricity traded with the main power grid, It means buying electricity from the main grid, whereas it means selling electricity to the main grid to gain extra profit; It indicates the upper limit of the power that the microgrid can purchase from the main grid or sell to the main grid. It indicates the lower limit of the microgrid purchasing electricity from the main grid or selling electricity to the main grid; and They respectively represent the lower and upper limits of the state of charge of the thermal energy storage device; and They respectively represent the lower and upper limits of the state of charge of the energy storage device.
[0024] Furthermore, step S5 specifically includes:
[0025] S51: normalize the historical operation data of the microgrid and initialize the energy price and the state of charge of the energy storage device;
[0026] S52: Initialize the policy network π of the safe energy scheduling agent φ , Evaluation Network Q θ and secure network weight parameters; in order to make the agent training process more stable, each network is equipped with a corresponding target network and initialized; the strategy network π φ The input is the microgrid environment state s t , the output is action at The probability density distribution of the microgrid environment state s t and action a t , evaluation network Q θ Output estimated cumulative expected reward, safety net Output estimated cumulative security cost expectation;
[0027] S53: Based on the data obtained in S51, the safe energy scheduling agent continuously interacts with the microgrid environment. According to the designed reward function and safety cost function, the immediate reward value and safety cost value fed back by the environment can be obtained, and the safe energy scheduling agent is trained based on this.
[0028] Furthermore, in step S6, after the safe energy scheduling agent outputs the action, the output action is checked and corrected, and then used to implement online low-carbon energy scheduling of the microgrid.
[0029] Furthermore, it also includes: during the operation of the microgrid, the interaction information of the intelligent agent environment is continuously collected, and the intelligent agent is updated online based on the collected information.
[0030] In the second aspect, a smart microgrid low-carbon energy dispatching system under security constraints is provided, including:
[0031] Sensing system: used to perceive microgrid environmental status information in real time;
[0032] Intelligent dispatch center: used to receive environmental status information sensed by the sensor system, and use the deployed safe energy dispatching agent to output dispatching instructions to dispatch energy units in the microgrid system;
[0033] Control system: Control the output power of energy units according to the dispatching instructions of the intelligent dispatching center;
[0034] Among them, the safe energy scheduling agent is obtained by the following method:
[0035] S1: Set safe operation constraints according to the microgrid structure and energy unit operating conditions;
[0036] S2: Calculate the cost according to the operation status of the microgrid and set the basic optimization target;
[0037] S3: Introduce carbon trading into microgrid operations and establish a joint electricity-carbon target;
[0038] S4: Convert the low-carbon optimization scheduling problem into a Markov decision process;
[0039] S5: Introduce a safety network to build a safe energy dispatching agent, and perform offline training of the safe energy dispatching agent based on the historical operation data of the microgrid and the Markov decision process.
[0040] Furthermore, the intelligent dispatching center includes:
[0041] Database module: used to store the environmental status information perceived by the sensor system at each moment and the scheduling instructions output by the intelligent agent;
[0042] Data processing module: normalizes environmental status information, calculates basic operating costs and carbon trading costs at each moment, and transforms the low-carbon optimization scheduling problem into a Markov decision process;
[0043] Reinforcement learning agent training module: performs offline training of the safe energy scheduling agent based on the data obtained by the data processing module;
[0044] Safety scheduling module: The safety energy scheduling agent obtained by the reinforcement learning agent training module outputs scheduling instructions, and checks and corrects the output scheduling instructions to ensure the safety of the scheduling instructions.
[0045] The present invention proposes a method and system for low-carbon energy scheduling of a smart microgrid under safety constraints, which has the following beneficial effects: The present invention constructs an electricity-carbon joint optimization scheduling target based on a carbon trading mechanism, incentivizes the system to operate in a low-carbon mode, and realizes low-carbon energy scheduling of smart microgrids. The present invention considers the state of charge constraints of energy storage equipment, energy trading constraints with the main power grid, and electricity / heat supply and demand balance constraints, and transforms the low-carbon energy scheduling problem into a Markov decision process. The present invention introduces a safety network to guide the policy network to learn in the direction of satisfying safety operation constraints during offline training, thereby realizing safe and low-carbon operation of smart microgrids. BRIEF DESCRIPTION OF THE DRAWINGS
[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0047] Figure 1 It is a flow chart of a low-carbon energy dispatching method for a smart microgrid under safety constraints provided by an embodiment of the present invention;
[0048] Figure 2 It is a schematic diagram of a smart microgrid low-carbon energy dispatching system under safe operation constraints provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0049] To make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described in detail below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other implementation methods obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.
[0050] like Figure 1 As shown, an embodiment of the present invention provides a low-carbon energy scheduling method for a smart microgrid under safety constraints, comprising the following steps:
[0051] S1: Set safe operation constraints according to the microgrid structure and energy unit operating conditions.
[0052] First, according to the structure of the microgrid, an input-output model of the energy unit is constructed; then safe operation constraints are set, such as power constraints of the energy unit, energy supply and demand balance constraints, power consumption constraints with the main grid, and energy storage device charge state constraints. In this embodiment, the microgrid is composed of energy units such as renewable energy power generation equipment, gas boilers, electric boilers, and energy storage equipment. It is worth noting that the energy units of the microgrid can be changed according to the actual architecture of the microgrid, and this implementation does not impose any restrictions. This embodiment takes into account safe operation constraints such as energy supply and demand balance constraints, power consumption constraints with the main grid, and energy storage device charge state constraints based on the structural composition of the microgrid.
[0053] S2: Calculate the cost based on the operation of the microgrid and set the basic optimization target. The basic optimization target includes the energy cost C generated by the transaction between the microgrid and the main grid. grid,t , energy cost generated by microgrid and natural gas grid transactions C gas,t and the investment and operation cost of the energy unit C invest,t .
[0054] S3: Introduce carbon trading into microgrid operations and establish joint electricity and carbon goals.
[0055] Constructing a joint electricity-carbon target L=C that comprehensively considers basic operating costs and carbon trading costs grid,t +C gas,t +C invest,t +C carbon,t , encouraging microgrids to operate in a low-carbon economic manner.
[0056] Carbon trading costs are specifically expressed as:
[0057]
[0058] In the formula, C carbon,t represents the carbon trading cost, represents the price of carbon trading, Indicates the amount of electricity purchased from the main power grid. Indicates the output power of the gas boiler. and They represent the unit power carbon emission factor and carbon quota coefficient of electricity purchased from the main grid, and They respectively represent the unit power carbon emission factor and carbon quota coefficient of the gas boiler.
[0059] S4: Convert the low-carbon optimization scheduling problem into a Markov decision process.
[0060] In this embodiment, the reward function is the opposite of the electricity-carbon joint optimization objective according to the microgrid operating status and the energy unit design state space and action space.
[0061] Specifically, in this embodiment, the state is expressed as Actions are represented by The reward function is expressed as r(s t ,a t )=-(C grid,t +C gas,t +C invest,t +C carbon,t ), where t represents the time, and They represent the price of selling electricity to the main grid, the price of purchasing electricity from the main grid, the price of purchasing natural gas, and the price of carbon trading. and They represent the heat energy demand and electricity energy demand of the user at time t, represents the power generation of the photovoltaic power generation equipment at time t, and They represent the charge states of the thermal energy storage device and the electrical energy storage device at time t, respectively. Indicates the charging and discharging power of the electrical energy storage device. and Represent the thermal power of thermal energy storage equipment, electric boiler and gas boiler respectively;
[0062] The safety cost function is introduced to represent the satisfaction of the safety operation constraints. The safety cost function includes the following safety cost functions: At least one of:
[0063]
[0064]
[0065] Where, [X] + =max(0,X), indicates that the safety operation constraints described are met, Indicates the amount of electricity traded with the main power grid, It means buying electricity from the main grid, whereas it means selling electricity to the main grid to gain extra profit; It indicates the upper limit of the power that the microgrid can purchase from the main grid or sell to the main grid. It indicates the lower limit of the microgrid purchasing electricity from the main grid or selling electricity to the main grid; and They respectively represent the lower and upper limits of the state of charge of the thermal energy storage device; and They represent the lower and upper limits of the state of charge of the energy storage device. By setting the upper and lower limits of the action space, the power constraints of the energy unit can be met.
[0066] S5: Introduce a safety network to build a safe energy dispatching agent, and perform offline training of the safe energy dispatching agent based on the historical operation data of the microgrid and the Markov decision process.
[0067] Specifically, step S5 includes:
[0068] S51: Normalize the microgrid historical operation data such as renewable energy generation power, user electricity / heat energy demand power, and initialize the energy price and charge state of the energy storage device.
[0069] S52: Initialize the policy network π of the safe energy scheduling agent φ , Evaluation Network Q θ and secure network weight parameters; in order to make the agent training process more stable, each network is equipped with a corresponding target network and initialized; the strategy network π φ The input is the microgrid environment state S t , the output is action A t The probability density distribution of the microgrid environment state s t and action A t , evaluation network Q θ Output estimated cumulative expected reward, safety net Output the estimated cumulative safety cost expectation. The introduction of the safety network guides the reinforcement learning agent to learn in the direction of satisfying the safety operation constraints. In this embodiment, each neural network has 2 hidden layers, and each hidden layer has 128 neurons.
[0070] S53: Based on the data obtained in S51, the safe energy dispatching agent continuously interacts with the microgrid environment, and the immediate reward value and safety cost value fed back by the environment can be obtained according to the designed reward function and safety cost function, and the safe energy dispatching agent is trained based on this to obtain the safe energy dispatching agent. During the training process, the target network is updated by soft update; the safety network can be updated by random policy gradient descent according to the designed safety cost value.
[0071] Design a safe action value function S π (s t ,a t ) is used to describe the strategy π, in state S t Execute action A t The expected value of the accumulated safety cost obtained.
[0072]
[0073] In the formula, γ represents the discount factor, c(s t ,a t ) indicates that in state S t Next, perform action a t The obtained safety cost value can be calculated by the safety cost function;
[0074] Designing a secure network To approximate the safe action value function, its loss function is:
[0075]
[0076] In the formula, Indicates state s t+1 The corresponding target safety cost state value, Represents the experience replay pool, Indicates that in state s t Next, perform action a t The resulting safety network output value;
[0077] To ensure that the agent can learn both rewards and safety constraints, a strategy network π is designed φ (·|s t )The updated objective function is specifically expressed as:
[0078]
[0079] In the formula, Q θ (s t ,a t ) means in state s t Next, perform action a tThe obtained evaluation network output value, α represents the temperature factor. In order to enable the intelligent agent to adapt to various safe operation constraints, multiple safety networks are constructed Use the corresponding security cost function Supports secure network Update of λ i (i=1,2,3,4) represents the Lagrange multiplier, d i (i=1, 2, 3, 4) represents the threshold value corresponding to the safety constraint.
[0080] S6: Use the safe energy dispatching agent to interact with the microgrid environment in real time to achieve online low-carbon energy dispatching of the microgrid.
[0081] Since the safety network guides the agent to learn during the training process, the agent has the ability to avoid violating constraints. In order to fully ensure the safety of the dispatching action, after the safe energy dispatching agent outputs the action, the output action is checked and corrected, and then used to realize the online low-carbon energy dispatch of the microgrid. In addition, during the operation of the microgrid, the interactive information of the agent environment is continuously collected, and the agent is updated online based on the collected information.
[0082] The above-mentioned embodiments propose a method and system for low-carbon energy scheduling of a smart microgrid under safety constraints, which has the following beneficial effects: the present invention constructs an electricity-carbon joint optimization scheduling target based on a carbon trading mechanism, incentivizes the system to operate in a low-carbon mode, and realizes low-carbon energy scheduling of smart microgrids. The present invention considers the state of charge constraints of energy storage equipment, energy trading constraints with the main power grid, and electricity / heat supply and demand balance constraints, and transforms the low-carbon energy scheduling problem into a Markov decision process. The present invention introduces a safety network to guide the policy network to learn in the direction of satisfying safety operation constraints during offline training, thereby realizing safe and low-carbon operation of smart microgrids.
[0083] This embodiment also provides a smart microgrid low-carbon energy dispatching system under security constraints, such as Figure 2 As shown, including:
[0084] Sensing system: used to perceive microgrid environmental status information in real time;
[0085] Intelligent dispatch center: used to receive environmental status information sensed by the sensor system, and use the deployed safe energy dispatch intelligent agent to output dispatch instructions to dispatch energy units in the microgrid system; wherein the safe energy dispatch intelligent agent is obtained by the method described in steps S1 to S5 in the above embodiment;
[0086] Control system: Controls the output power of energy units according to the dispatching instructions of the intelligent dispatching center.
[0087] Specifically, the intelligent dispatching center includes:
[0088] Database module: used to store the environmental status information perceived by the sensor system at each moment and the scheduling instructions output by the intelligent agent;
[0089] Data processing module: normalizes environmental status information, calculates basic operating costs and carbon trading costs at each moment, and transforms the low-carbon optimization scheduling problem into a Markov decision process;
[0090] Reinforcement learning agent training module: performs offline training of the safe energy scheduling agent based on the data obtained by the data processing module;
[0091] Safety scheduling module: The safety energy scheduling agent obtained by the reinforcement learning agent training module outputs scheduling instructions, and checks and corrects the output scheduling instructions to ensure the safety of the scheduling instructions.
[0092] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.
[0093] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.
Claims
1. A low-carbon energy dispatching method for a smart microgrid under security constraints, characterized in that: The steps include: S1: Set safe operation constraints according to the microgrid structure and energy unit operating conditions; S2: Calculate the cost according to the operation status of the microgrid and set the basic optimization target; S3: Introduce carbon trading into microgrid operations and establish a joint electricity-carbon target; S4: Convert the low-carbon optimization scheduling problem into a Markov decision process; S5: Introduce a safety network to build a safe energy dispatching agent, and perform offline training of the safe energy dispatching agent based on the historical operation data of the microgrid and the Markov decision process; S6: Use the safe energy dispatching agent to interact with the microgrid environment in real time to achieve online low-carbon energy dispatching of the microgrid.
2. The method for low-carbon energy dispatching of a smart microgrid under security constraints according to claim 1 is characterized in that: In step S1, the safe operation constraints include at least one of the power constraints of the energy unit, the energy supply and demand balance constraints, the power transaction constraints with the main power grid and the charge state constraints of the energy storage device.
3. The method for low-carbon energy dispatching of a smart microgrid under security constraints according to claim 1 is characterized in that: In step S2, the basic optimization objectives include the energy cost generated by the microgrid's transactions with the main grid, the energy cost generated by the microgrid's transactions with the natural gas grid, and the investment and operation costs of the energy units.
4. The method for low-carbon energy dispatching of a smart microgrid under security constraints according to claim 1 is characterized in that: In step S3, a carbon trading mechanism is introduced to quantify the carbon trading cost, which is specifically expressed as: In the formula, C carbon,t represents the carbon trading cost, represents the price of carbon trading, Indicates the amount of electricity purchased from the main power grid. Indicates the output power of the gas boiler. and They represent the unit power carbon emission factor and carbon quota coefficient of electricity purchased from the main grid, and They respectively represent the unit power carbon emission factor and carbon quota coefficient of the gas boiler.
5. The method for low-carbon energy dispatching of a smart microgrid under security constraints according to claim 1 is characterized in that: In step S4, the state is represented as Actions are represented by The reward function is expressed as r(s t ,a t )=-(C grid,t +C gas,t +C invest,t +C carbon,t ), where t represents the time, and They represent the price of selling electricity to the main grid, the price of purchasing electricity from the main grid, the price of purchasing natural gas, and the price of carbon trading. and They represent the heat energy demand and electricity energy demand of the user at time t, represents the power generation of the photovoltaic power generation equipment at time t, and They represent the charge states of the thermal energy storage device and the electrical energy storage device at time t, respectively. Indicates the charging and discharging power of the electrical energy storage device. and Represent the thermal power of thermal energy storage equipment, electric boiler and gas boiler respectively, C grid,t represents the energy cost generated by the microgrid and the main grid, C gas,t represents the energy cost generated by the transaction between the microgrid and the natural gas grid, C invest,t represents the investment and operation cost of the energy unit, C carbon,t represents the carbon trading cost; The safety cost function is introduced to represent the satisfaction of the safety operation constraints. The safety cost function includes the following safety cost functions: At least one of: Where, [x] + =max(0,x), indicates that the safety operation constraints described are met, Indicates the amount of electricity traded with the main power grid, It means buying electricity from the main grid, whereas it means selling electricity to the main grid to gain extra profit; It indicates the upper limit of the power that the microgrid can purchase from the main grid or sell to the main grid. It indicates the lower limit of the microgrid purchasing electricity from the main grid or selling electricity to the main grid; and They respectively represent the lower and upper limits of the state of charge of the thermal energy storage device; and They respectively represent the lower and upper limits of the state of charge of the energy storage device.
6. The method for low-carbon energy dispatching of a smart microgrid under security constraints according to claim 1, characterized in that: Step S5 specifically includes: S51: normalize the historical operation data of the microgrid and initialize the energy price and the state of charge of the energy storage device; S52: Initialize the policy network π of the safe energy scheduling agent φ , Evaluation Network Q θ and secure network weight parameters; in order to make the agent training process more stable, each network is equipped with a corresponding target network and initialized; the strategy network π φ The input is the microgrid environment state s t , the output is action a t The probability density distribution of the microgrid environment state s t and action a t , evaluation network Q θ Output estimated cumulative expected reward, safety net Output estimated cumulative security cost expectation; S53: Based on the data obtained in S51, the safe energy scheduling agent continuously interacts with the microgrid environment. According to the designed reward function and safety cost function, the immediate reward value and safety cost value fed back by the environment can be obtained, and the safe energy scheduling agent is trained based on this.
7. The method for low-carbon energy dispatching of a smart microgrid under security constraints according to claim 1 is characterized in that: In step S6, after the safe energy scheduling agent outputs the action, the output action is checked and corrected, and then used to implement online low-carbon energy scheduling of the microgrid.
8. The method for low-carbon energy dispatching of a smart microgrid under security constraints according to claim 1, characterized in that: Also includes: During the operation of the microgrid, the interactive information of the agent environment is continuously collected, and the agent is updated online based on the collected information.
9. A smart microgrid low-carbon energy dispatching system under security constraints, characterized in that: include: Sensing system: used to perceive microgrid environmental status information in real time; Intelligent dispatch center: used to receive environmental status information sensed by the sensor system, and use the deployed safe energy dispatching agent to output dispatching instructions to dispatch energy units in the microgrid system; Control system: Control the output power of energy units according to the dispatching instructions of the intelligent dispatching center; Among them, the safe energy scheduling agent is obtained by the following method: S1: Set safe operation constraints according to the microgrid structure and energy unit operating conditions; S2: Calculate the cost according to the operation status of the microgrid and set the basic optimization target; S3: Introduce carbon trading into microgrid operations and establish a joint electricity-carbon target; S4: Convert the low-carbon optimization scheduling problem into a Markov decision process; S5: Introduce a safety network to build a safe energy dispatching agent, and perform offline training of the safe energy dispatching agent based on the historical operation data of the microgrid and the Markov decision process.
10. The smart microgrid low-carbon energy dispatching system under security constraints according to claim 9 is characterized in that: The intelligent dispatching center comprises: Database module: used to store the environmental status information perceived by the sensor system at each moment and the scheduling instructions output by the intelligent agent; Data processing module: normalizes environmental status information, calculates basic operating costs and carbon trading costs at each moment, and transforms the low-carbon optimization scheduling problem into a Markov decision process; Reinforcement learning agent training module: performs offline training of the safe energy scheduling agent based on the data obtained by the data processing module; Safety scheduling module: The safety energy scheduling agent obtained by the reinforcement learning agent training module outputs scheduling instructions, and checks and corrects the output scheduling instructions to ensure the safety of the scheduling instructions.
Citation Information
Cited By
Low-carbon dispatching control method and device for power system, terminal equipment and storage medium
CN119726827A
A low-carbon dispatching control method and device for a power system, a terminal device, and a storage medium
CN119726827B
Electric vehicle low-carbon scheduling method and system based on two-stage diffusion strategy and multi-agent reinforcement learning
CN121543990A
