CN-121998231-A - System and method for predicting electric power carbon emission intensity and making auxiliary decisions for carbon transaction

Abstract

The invention discloses a system and a method for predicting electric power carbon emission intensity and making carbon transaction assistance decisions, and belongs to the technical field of intelligent energy and carbon asset management. The system comprises a data preprocessing module, a carbon emission intensity prediction module and a carbon transaction auxiliary decision-making module. The method comprises the steps of processing multi-source data from electric power, weather and carbon markets, predicting the carbon emission intensity of future electric power by utilizing a trained long-term and short-term memory network model, performing carbon price multi-scenario simulation based on integrated learning and Monte Carlo simulation, evaluating transaction strategy risks by adopting conditional risk values, integrating prediction, simulation and risk evaluation results, and generating a quantized transaction strategy by solving a multi-objective optimization problem aiming at maximizing benefits and minimizing the conditional risk values. The invention realizes high-precision short-term prediction of carbon emission intensity, builds a prediction-decision integrated model driving frame on the basis, effectively solves the pain point of carbon transaction depending on experience decision, and can assist users to conduct risk-controllable intelligent carbon asset transaction.

Inventors

• LIU XIULIANG

Assignees

• 刘秀良

Dates

Publication Date

20260508

Application Date

20251207

Claims (18)

1. 1. An electrical carbon emission intensity prediction and carbon trade aid decision-making system, comprising: The data preprocessing module is used for receiving and processing multi-source heterogeneous data from a power system, a meteorological environment and a carbon market; The carbon emission intensity prediction module is connected with the data preprocessing module and used for predicting the electric power carbon emission intensity of a specified time period in the future through a trained long-period and short-period memory network model based on the processed data; The carbon transaction auxiliary decision-making module is connected with the carbon emission intensity prediction module and is used for generating a quantitative transaction strategy at least according to the predicted electric power carbon emission intensity; wherein the carbon transaction assistance decision making module comprises: the carbon price simulation sub-module is used for carrying out multi-scenario simulation of carbon price based on historical data and market indexes; The risk assessment sub-module is used for assessing the risk of the transaction strategy based on the conditional risk value model; And the strategy optimization engine is used for integrating the carbon emission intensity prediction result, the carbon price simulation scene and the risk assessment result, and outputting transaction strategy recommendation comprising transaction opportunity, quantity and risk gain indexes by solving a multi-objective optimization function aiming at maximizing expected benefits and minimizing conditional risk values.
2. 2. The system of claim 1, wherein the data preprocessing module is specifically configured to perform data cleaning, missing value filling, and Z-score normalization processing on the multi-source heterogeneous data, and screen out a feature variable set with highest association with carbon intensity and carbon price by using a mutual information method.
3. 3. The system of claim 1, wherein the long-term memory network model employed by the carbon emission intensity prediction module is a stacked structure comprising two LSTM hidden layers, and a Dropout layer is disposed behind the hidden layers.
4. 4. The system of claim 3, wherein the input characteristic variables of the long-short term memory network model include at least historical carbon emission intensity sequences, thermal generator set output data, wind and photovoltaic generator set output data, grid load data, temperature data, and wind speed data.
5. 5. The system of claim 1, wherein the carbon number simulation sub-module employs a Stacking integrated learning framework, the base learner layer comprises a LASSO regression model and a gradient lifting decision tree model, and the meta learner layer uses a linear regression model to fuse the outputs of the base learner to obtain the predicted value of the carbon number point.
6. 6. The system of claim 5, wherein the carbon number simulation submodule further comprises a monte carlo simulation unit for randomly sampling based on a residual distribution of the carbon number point predictions, generating a plurality of simulation paths of future carbon numbers, and defining three market scenarios of benchmark, optimistic and pessimistic based on a statistical distribution of the simulation paths.
7. 7. The system of claim 1, wherein the risk assessment submodule is configured to construct a damage distribution of the transaction strategy to be assessed based on the plurality of simulated paths generated by the carbon price simulation submodule and calculate a conditional risk value at a 95% confidence level based on the damage distribution.
8. 8. The system of claim 1, wherein the constraints of the multi-objective optimization function include a single transaction upper cost constraint, a total inventory upper constraint, and a total transaction budget constraint.
9. 9. The system of claim 1, wherein the policy optimization engine is configured to solve the multi-objective optimization function using a non-dominant ordered genetic algorithm with elite policy to obtain a pareto optimal policy set, and select a final recommended policy from the set based on a criterion that a summer ratio is highest.
10. 10. The electric power carbon emission intensity prediction and carbon transaction auxiliary decision-making method is characterized by comprising the following steps of: s1, collecting multi-source heterogeneous data from a power system, a meteorological environment and a carbon market, and preprocessing the data including cleaning, normalization and feature screening; S2, inputting the preprocessed data into a pre-trained long-short-period memory network prediction model with a stacked structure to obtain an electric power carbon emission intensity predicted value for 24 hours in the future; S3, carrying out multi-scenario simulation of carbon prices by adopting a method combining Stacking integrated learning and Monte Carlo simulation based on a historical carbon price sequence, energy price and macroscopic economic indexes; S4, aiming at a potential transaction strategy, constructing damage distribution of the potential transaction strategy based on a multi-scenario simulation result of the carbon price, and calculating conditional risk value of the potential transaction strategy under a preset confidence level; And S5, taking the predicted value of the carbon emission intensity, the simulation scene of the carbon price and the evaluation result of the conditional risk value as inputs, and generating an optimal transaction strategy by solving a multi-objective optimization problem which aims at maximizing expected benefits and minimizing the conditional risk value.
11. 11. The method of claim 10, wherein in step S1, the feature screening employs a mutual information method, and the screened key features include historical carbon intensity, thermal power output, new energy output, system load, wind speed and illumination intensity.
12. 12. The method of claim 10, wherein the training process of the long-term and short-term memory network prediction model in step S2 includes dividing a historical data set into a training set, a verification set and a test set according to a ratio of 7:2:1, minimizing a mean square error loss function by using the training set and an Adam optimizer, and triggering an early-stop mechanism when the loss function value on the verification set does not drop in 10 epochs in succession.
13. 13. The method according to claim 10, wherein step S3 specifically comprises: s31, generating a carbon price datum point prediction by using a Stacking integrated model comprising a linear base learner and a nonlinear base learner; S32, generating more than 10000 future carbon number paths through Monte Carlo simulation based on residual error distribution predicted by the datum point; And S33, determining a carbon price prediction interval according to 97.5% and 2.5% quantiles of the future carbon price path, and defining benchmark, optimistic and pessimistic scenes according to the carbon price prediction interval.
14. 14. The method according to claim 10, characterized in that in step S4 the conditional risk value is calculated in particular by calculating the average loss in the worst case of the last 5% of the profit-and-loss distribution at a confidence level of 95%.
15. 15. The method according to claim 10, wherein in step S5, the mathematical model of the multi-objective optimization problem is: Maximize: [R(x), -CVaR_α(x)] Subject to: C_transaction(x) ≤ C_max, Position(x) ≤ P_max, Capital(x) ≤ B_total Where x is the decision variable, R (x) is the expected yield function, CVaR _α (x) is the conditional risk cost function at confidence level α, C_ transaction (x) is the total transaction cost, C_max is the upper cost limit, position (x) is the total holding capacity, P_max is the upper holding capacity limit, catal (x) is the occupied funds, and B_total is the total budget.
16. 16. The method of claim 10, wherein the optimal transaction strategy generated in step S5 includes recommendation output information including a specific date and hour window for transaction execution, a precise number of carbon quota transactions, expected annual rate of return, a summer rate, a daily risk value at a 95% confidence level, and a conditional risk value.
17. 17. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the method of any of claims 10 to 16 when the computer program is executed.
18. 18. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method according to any of claims 10 to 16.

Description

System and method for predicting electric power carbon emission intensity and making auxiliary decisions for carbon transaction Technical Field The invention belongs to the technical field of intelligent energy and carbon emission management, and particularly relates to a decision support system and method for fusing data-driven prediction and quantitative financial models, in particular to a system and method for predicting electric power carbon emission intensity and assisting in decision making of carbon market transaction. Background In the context of global active management of climate change and the deep advancement of "two-carbon" strategy in china, the power industry is a key sector of carbon emissions, and low-carbon transformation is of paramount importance. The formal operation of the national carbon emission right trading market makes it necessary for the power generation enterprises not only to manage their own carbon emissions, but also to actively participate in the carbon market trading to optimize assets, control costs and obtain benefits. However, current electrical power systems have significantly short plates in terms of real-time, accurate sensing of carbon emissions. The traditional carbon emission accounting is based on post statistics and emission factor method, and can not provide dynamic and high-precision prediction of carbon emission intensity for hours or days in the future. This makes it difficult for an enterprise to pre-judge its own carbon emission performance risk in advance, nor to pre-incorporate the carbon emission signal into the operational decision. At the carbon trade decision level, market participants currently rely on historical experience, qualitative analysis and simple rules for trade, and a systematic and model-driven quantitative decision tool is lacking. Carbon price is influenced by multiple complex factors such as energy price, policy, macroscopic economy and the like, volatility is large, and traditional experience is difficult to effectively capture market rules and manage risks. Although some researches on carbon price prediction or trading strategies exist, most of carbon emission prediction and trading decisions are split, and an integrated intelligent decision closed loop from front-end carbon situation awareness to back-end risk and benefit optimization cannot be constructed. Therefore, a technical scheme for realizing high-precision short-term prediction of the electric power carbon emission intensity and performing intelligent and quantitative carbon transaction auxiliary decision-making based on the prediction result and combining an advanced financial risk model is urgently needed, so that the core pain point which depends on experience and rough decision is solved. Disclosure of Invention Object of the invention The invention aims to overcome the defects of the prior art and provides a system and a method for predicting the carbon emission intensity of electric power and assisting in decision making of carbon transaction. The core purpose is that: Short-term (e.g., 24 hours in the future) high-precision prediction of regional or node-level electric power carbon emission intensity is realized, and a precise mapping relation of 'electric power system operation behavior-carbon emission intensity' is established. The carbon emission intensity prediction signal, the carbon price market simulation and the financial risk management tool are deeply fused to construct an integrated decision frame of prediction, simulation, evaluation and optimization. And outputting quantized, executable and risk-controllable carbon transaction strategy suggestions, assisting market participants such as power generation enterprises to change from experience decision-making to data and model driven decision-making, and improving the operation benefit and risk management capability of the market participants in the carbon market. (II) technical scheme In order to achieve the above purpose, the invention adopts the following technical scheme: an electrical carbon emission intensity prediction and carbon trade aid decision making system comprising: The data preprocessing module is used for receiving and processing multi-source heterogeneous data from a power system, a meteorological environment and a carbon market; The carbon emission intensity prediction module is connected with the data preprocessing module and is used for predicting the electric power carbon emission intensity of a specified time period in the future through a trained long-short-term memory network (LSTM) model based on the processed data; The carbon transaction auxiliary decision-making module is connected with the carbon emission intensity prediction module and is used for generating a quantitative transaction strategy at least according to the predicted electric power carbon emission intensity; wherein the carbon transaction assistance decision making module comprises: the carbon price simulation sub-