Electric power spot strategy processing method based on intelligent agent and computer equipment

Abstract

The invention relates to the technical field of electric power markets, and discloses an intelligent agent-based electric power spot strategy processing method and computer equipment, and the method comprises the steps: obtaining the current individual state data of a target individual and the current environment state data of an electric power spot market; inputting the current individual state data and the current environment state data into a first target agent to generate a plurality of candidate strategies; wherein the candidate strategies correspond to specified strategy factors; inputting the current environment state data into a second target agent to determine an acceptance probability corresponding to the specified strategy factor; and screening the plurality of candidate strategies according to the acceptance probability to obtain a target strategy for the target individual to carry out electric power spot transaction on the current operation day. According to the method, the scientificity, the accuracy and the efficiency of strategy generation and screening are greatly improved, and the target individual can achieve the maximum income in the electric power spot market.

Claims

1. A method for processing power spot strategy of an intelligent agent, characterized in that: The method comprises: Acquire current individual state data of the target individual and current environmental state data of the electricity spot market, wherein the current individual state data is used to characterize the predicted power generation of the target individual on the current operation day, and the current environmental state data is used to characterize the predicted total load and total unit output in the electricity spot market on the current operation day; Inputting the current individual state data and the current environment state data into the first target intelligent agent to generate a plurality of candidate strategies; wherein the candidate strategies correspond to designated strategy factors; Inputting the current environment state data into the second target intelligent agent to determine the acceptance probability corresponding to the specified strategy factor; The plurality of candidate strategies are screened according to the acceptance probability to obtain a target strategy for the target individual to conduct spot electricity trading on the current operating day.

2. The power spot strategy processing method according to claim 1, characterized in that: The designated strategic factor is at least one of a load factor of the electricity spot market, a wind power output factor, a photovoltaic power output factor, a power transmission factor, and a power generation factor of a target individual.

3. The power spot strategy processing method according to claim 2, characterized in that: The first target intelligent agent includes a plurality of simple strategy generation layers and a reference strategy generation layer, wherein the simple strategy generation layers correspond one to one with the specified strategy factors.

4. The power spot strategy processing method according to claim 3 is characterized in that: The plurality of simple policy generation layers include a first simple policy generation layer, a second simple policy generation layer, a third simple policy generation layer and a fourth simple policy generation layer; The step of inputting the current individual state data and the current environment state data into the first target intelligent agent to generate a plurality of candidate strategies comprises: Inputting the predicted total power generation of all the units in the current individual state data and the current predicted total load in the current environmental state data into the first simple strategy generation layer to generate a first candidate strategy corresponding to the load factor; Inputting the predicted wind power generation of the wind turbine group in the current individual state data and the predicted total wind power output of the wind turbine group in the current environmental state data into the second simple strategy generation layer to generate a second candidate strategy corresponding to the wind power output factor; Inputting the predicted photovoltaic power generation of the photovoltaic group in the current individual state data and the predicted total photovoltaic output of the photovoltaic group in the current environmental state data into the third simple strategy generation layer to generate a third candidate strategy corresponding to the photovoltaic output factor; Inputting the predicted individual outgoing power of all the units in the current individual state data and the predicted total outgoing power of all the units in the current environmental state data into a fourth simple strategy generation layer to generate a fourth candidate strategy corresponding to the outgoing power factor; Inputting the predicted total power generation of all the units in the current individual state data and the installed capacity of the target unit into the benchmark strategy generation layer, and determining the ratio between the predicted total power generation and the installed capacity, so as to generate a benchmark candidate strategy corresponding to the power generation factor according to the ratio; The multiple candidate strategies include the first candidate strategy, the second candidate strategy, the third candidate strategy, the fourth candidate strategy and the benchmark candidate strategy.

5. The power spot strategy processing method according to claim 3 or 4, characterized in that: The simple strategy generation layer is a neural network constructed based on the DDPG algorithm, and each of the simple strategy generation layers includes an actor neural network using a sigmoid activation function and a critic neural network using a tanh activation function, wherein the actor neural network is used to generate multiple simple strategies corresponding to the specified strategy factors, and the critic neural network is used to determine the candidate strategies corresponding to the corresponding specified strategy factors according to the benefits of the multiple simple strategies.

6. The power spot strategy processing method according to claim 1, characterized in that: The second target intelligent agent includes an actor neural network, and the actor neural network adopts a DQN algorithm and a sigmoid activation function; the inputting the current environment state data into the second target intelligent agent to determine the acceptance probability corresponding to the specified strategy factor includes: Inputting the current environment state data into the actor neural network of the second target agent to score the suitability of each of the specified strategy factors under the current environment state data based on the DQN algorithm; The suitability score is normalized according to the sigmoid activation function to obtain the acceptance probability corresponding to each of the specified strategy factors.

7. The power spot strategy processing method according to claim 1, characterized in that: The screening of the plurality of candidate strategies according to the acceptance probability to obtain a target strategy for the target individual to conduct spot electricity trading on the current operation day includes: A target policy factor corresponding to the maximum value of the acceptance probability is determined, and a candidate policy corresponding to the target policy factor is selected from the multiple candidate policies as the target policy.

8. The power spot strategy processing method according to claim 1, characterized in that: The training process of the first target agent includes: Obtaining historical strategy benefits and historical benchmark benefits for historical operating days, wherein the historical strategy benefits are the actual benefits of the target individual on the historical operating days, and the historical benchmark benefits are the benefits obtained by the target individual from electricity spot transactions according to the historical unit forecast power generation on the historical operating days; An original reward is determined according to the difference between the historical strategy return and the historical benchmark return, and a first initial agent is trained according to the original reward to obtain the first target agent.

9. The power spot strategy processing method according to claim 8, characterized in that: The training process of the second target agent includes: Determining a continuous loss indicator of earnings of the target individual according to the original reward; When the continuous loss index of the income reaches a preset limit, the original reward is proportionally adjusted according to the preference adjustment hyperparameter of the target individual to obtain an adjusted reward value; Training the second initial agent according to the adjusted reward value to obtain the second target agent; The profit continuous loss indicator includes at least one of the continuous loss days, the historical maximum continuous loss days, the continuous loss amount and the historical maximum continuous loss amount.

10. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the electricity spot strategy processing method according to any one of claims 1 to 9 by executing the computer instructions.

Images