CN-121998277-A - Intelligent dispatching method, system, equipment and medium for distributed energy transaction based on predictive algorithm

CN121998277ACN 121998277 ACN121998277 ACN 121998277ACN-121998277-A

Abstract

The invention discloses a prediction algorithm-based distributed energy transaction intelligent scheduling method, a prediction algorithm-based distributed energy transaction intelligent scheduling system, a prediction algorithm-based distributed energy transaction intelligent scheduling device and a prediction algorithm-based distributed energy transaction intelligent scheduling medium, and relates to the technical field of distributed energy transaction scheduling. According to the invention, the output prediction precision is improved by combining multi-source data fusion with a probability prediction technology, and dynamic balance of benefits and risks is realized by optimizing a self-adaptive transaction strategy, so that a distributed energy transaction system can autonomously balance economic benefits and operation risks in a complex market environment, and the default probability is controlled by dynamic quota setting and a multi-stage early warning mechanism.

Inventors

LIU YAN
HE CHUNXIAO
HE GUOXIN
ZHAO LIN
DONG JIAN
GUO SHANGMIN
CHEN RUOJING
CHENG MENGZENG
LIU WEI
Meng Jingrong
LI YILIANG

Assignees

国网辽宁省电力有限公司经济技术研究院
上海交通大学四川研究院

Dates

Publication Date: 20260508
Application Date: 20251128

Claims (10)

1. A distributed energy transaction intelligent scheduling method based on a prediction algorithm is characterized by comprising the following steps of, Establishing a multi-source data acquisition system to acquire operation data of the distributed energy system and preprocessing the operation data to obtain an operation data set; performing feature extraction on the operation data set by adopting a multi-dimensional feature extraction project and combining a feature fusion algorithm to obtain a fusion feature sequence; a multi-step output prediction model is established by adopting a long-short memory network, and an output prediction interval is output by inputting a fusion characteristic sequence; Establishing a transaction strategy optimization model by adopting a reinforcement learning algorithm according to the output prediction interval, and outputting an optimal transaction strategy; Establishing a risk value calculation formula to set transaction limits according to the output optimal transaction strategy; And establishing a real-time risk monitoring model to calculate transaction risk indexes, and carrying out early warning according to the transaction risk indexes.
2. The intelligent scheduling method for distributed energy transaction based on predictive algorithm as set forth in claim 1, wherein said establishing a multi-source data acquisition system for acquiring and preprocessing operational data of the distributed energy system to obtain an operational data set comprises: establishing a multi-source data acquisition system to acquire operation data of a distributed energy system; performing data cleaning on the operation data of the distributed energy system; and carrying out normalization processing on the operation data of the data-cleaned distributed energy system to obtain an operation data set.
3. The intelligent scheduling method for distributed energy transaction based on predictive algorithm as set forth in claim 2, wherein the feature extraction and feature fusion algorithm for the operation data set by using the multidimensional feature extraction engineering to obtain a fusion feature sequence comprises the following steps: performing feature extraction on the operation data set by adopting a multi-dimensional feature extraction project; establishing an attention weight calculation formula to calculate the weight of the extracted features; and combining a feature fusion algorithm to fuse the calculated feature weights to obtain a fusion feature sequence.
4. The intelligent scheduling method for distributed energy transaction based on predictive algorithm as set forth in claim 3, wherein said establishing a multi-step output prediction model by using a long-short time memory network, inputting a fusion characteristic sequence and outputting an output prediction interval comprises: Establishing a multi-step output prediction model by adopting a long-short memory network; Building a loss function to train a multi-step growth force prediction model; And (3) according to the trained multi-step growth output prediction model, inputting the fusion characteristic sequence and outputting an output prediction interval.
5. The intelligent scheduling method for distributed energy trading based on predictive algorithm as set forth in claim 4, wherein the method for establishing a trading strategy optimization model by reinforcement learning algorithm according to the output prediction interval and outputting an optimal trading strategy comprises the steps of: According to the output prediction interval, a transaction strategy optimization model is established by adopting a reinforcement learning algorithm, and the expression is as follows: , Wherein, the For an instant prize at time t, For the electricity price of the first period, For the amount of transaction power in the first time period, Is the actual power generation amount in the first period, To make the price of the violation unit, paying the violation cost when the transaction electric quantity exceeds the actual electric quantity, The representation takes the absolute value of the value, Reflects decision maker risk preferences for risk aversion coefficients, In order to invest in the standard deviation of the portfolio, Indicating the total revenue of the transaction, Indicating the total cost of the breach, Representing a risk penalty; Training a transaction strategy optimization model by adopting an experience playback mechanism; And calculating the mean square error to update the transaction strategy optimization model, wherein the expression is as follows: , Wherein, the For the Q-network loss, In order to evaluate the network output, For the state of the kth sample, The action selected for the kth sample, Is the target Q value for the kth sample.
6. The intelligent scheduling method for distributed energy transaction based on predictive algorithm as set forth in claim 5, wherein said establishing a risk value calculation formula and setting a transaction limit according to the output optimal transaction strategy comprises: According to the output optimal transaction strategy, a risk value calculation formula is established through a random simulation method, wherein the expression is as follows: , , Wherein, the For a 95% confidence level risk value, For a 5% quantile of benefit, For a 95% confidence level conditional risk value, The i value after the profit value is sequenced from small to large; calculating the risk value of the transaction strategy through a risk value calculation formula; setting a transaction limit for energy transaction according to the calculated risk value, wherein the expression is as follows: , Wherein, the Representing the average benefit.
7. The intelligent scheduling method for distributed energy transaction based on predictive algorithm as set forth in claim 6, wherein said establishing a real-time risk monitoring model calculates transaction risk indicators, and pre-warning is performed according to the transaction risk indicators, comprising: According to the transaction strategy output process, a real-time risk monitoring model is established, and the expression is: , Wherein, the In order to predict the percentage of deviation, For the actual output force to be applied, In order to predict the median value of the data, Is rated capacity; Calculating transaction risk indexes according to the real-time risk monitoring model; and early warning is carried out according to the transaction risk index, and an emergency response mechanism is started.
8. The intelligent dispatching system for distributed energy transaction based on the predictive algorithm is applied to the intelligent dispatching method for distributed energy transaction based on the predictive algorithm as claimed in any one of claims 1 to 7, and is characterized by comprising an acquisition module, a feature extraction module, an output prediction module, a strategy optimization module, a quota setting module and a risk early warning module; the acquisition module is used for establishing a multi-source data acquisition system to acquire the operation data of the distributed energy system and preprocessing the operation data to obtain an operation data set; The feature extraction module adopts multidimensional feature extraction engineering to perform feature extraction on the operation data set and combines a feature fusion algorithm to obtain a fusion feature sequence; the output prediction module establishes a multi-step output prediction model by adopting a long-short memory network, inputs the fusion characteristic sequence and outputs an output prediction interval; the strategy optimization module establishes a transaction strategy optimization model by adopting a reinforcement learning algorithm according to the output prediction interval, and outputs an optimal transaction strategy; The quota setting module establishes a risk value calculation formula to set transaction quota according to the output optimal transaction strategy; The risk early warning module is used for establishing a real-time risk monitoring model to calculate transaction risk indexes and carrying out early warning according to the transaction risk indexes.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of a predictive algorithm based distributed energy transaction intelligent scheduling method according to any one of claims 1 to 7.
10. A computer readable storage medium having stored thereon a computer program, characterized in that the computer program when executed by a processor implements the steps of a predictive algorithm based distributed energy transaction intelligent scheduling method according to any one of claims 1 to 7.

Description

Intelligent dispatching method, system, equipment and medium for distributed energy transaction based on predictive algorithm Technical Field The invention relates to the technical field of distributed energy transaction scheduling, in particular to a distributed energy transaction intelligent scheduling method, system, equipment and medium based on a prediction algorithm. Background With the continuous increase of permeability of distributed energy sources in an electric power system, the intermittence and fluctuation of the output of the distributed energy sources provide great challenges for electric power market trading. The existing distributed energy transaction system mainly depends on a simple statistical prediction method, and has the following defects: Traditional prediction methods rely on a single data source resulting in poor accuracy. The existing distributed energy output prediction mainly adopts a time sequence analysis method, such as an autoregressive moving average model and an exponential smoothing method. These methods lack the ability to extract features at different time scales and cannot capture the multi-scale fluctuation law of distributed energy output. Uncertainty is not considered by the policy optimization model, which results in high transaction risk. The existing energy transaction scheduling method mostly adopts a deterministic optimization model, and takes a predicted value as a determined quantity to carry out transaction planning. Fixed trading plans may result in a severe deficit when market prices fluctuate dramatically. The existing method lacks modeling of prediction error probability distribution, can not quantify transaction risks, and can not formulate risk avoidance strategies. The lack of adaptive capability in fixed transaction strategies results in less than optimal revenue. Existing trading systems mostly employ fixed policies based on rules, such as simple threshold policies for low price purchases and high price sales. Market characteristic differences of different time periods are not considered, and targeted optimization is not carried out on typical time periods such as morning and evening peaks, afternoon valleys and the like. And when the prediction error is large, the aggressive strategy is still executed, and the risk control capability is insufficient. Disclosure of Invention The present invention has been made in view of the above-described problems occurring in the prior art. Therefore, the invention solves the technical problem of how to realize the prediction of the output of the distributed energy source by the multi-source data fusion and depth feature extraction technology in the distributed energy source transaction environment. And constructing a probability prediction model to quantify the prediction uncertainty, designing a self-adaptive transaction strategy optimization algorithm, and realizing the maximization of transaction income on the premise of controllable risk. In order to solve the technical problems, the invention provides a distributed energy transaction intelligent scheduling method based on a prediction algorithm, which comprises the following steps of, Establishing a multi-source data acquisition system to acquire operation data of the distributed energy system and preprocessing the operation data to obtain an operation data set; performing feature extraction on the operation data set by adopting a multi-dimensional feature extraction project and combining a feature fusion algorithm to obtain a fusion feature sequence; a multi-step output prediction model is established by adopting a long-short memory network, and an output prediction interval is output by inputting a fusion characteristic sequence; Establishing a transaction strategy optimization model by adopting a reinforcement learning algorithm according to the output prediction interval, and outputting an optimal transaction strategy; Establishing a risk value calculation formula to set transaction limits according to the output optimal transaction strategy; And establishing a real-time risk monitoring model to calculate transaction risk indexes, and carrying out early warning according to the transaction risk indexes. As a preferable scheme of the intelligent scheduling method for distributed energy transaction based on the prediction algorithm, the invention adopts multidimensional feature extraction engineering to perform feature extraction on an operation data set and combine a feature fusion algorithm to obtain a fusion feature sequence, and comprises the following steps: performing feature extraction on the operation data set by adopting a multi-dimensional feature extraction project; establishing an attention weight calculation formula to calculate the weight of the extracted features; and combining a feature fusion algorithm to fuse the calculated feature weights to obtain a fusion feature sequence. The invention makes the time sequence fluctuation rule, the pe