CN-121981307-A - MOBO algorithm driven key enterprise energy-saving carbon-reduction multi-target auxiliary decision-making method, system, equipment and medium

CN121981307ACN 121981307 ACN121981307 ACN 121981307ACN-121981307-A

Abstract

The invention discloses a MOBO algorithm-driven key enterprise energy-saving carbon-reduction multi-target auxiliary decision-making method, a system, equipment and a medium, which belong to the technical field of industrial energy conservation and emission reduction. The method responds to the time-sharing electricity price, the carbon factor and the load state, solves the problem of multi-target splitting of time-sharing electricity consumption, adopts MOBO algorithm to model the target function through Gaussian process regression model, can quickly respond to external condition mutation, effectively processes complex constraint, remarkably improves the performance of multi-target optimization in a dynamic energy-saving carbon reduction scene, and has higher convergence rate and better optimization effect.

Inventors

CHEN JULONG
WANG WEI
LUO NING
ZHU YONGQING
WANG YUXIANG
YIN JIA
LIN CHAO
WANG BIN
ZHANG YU
MOU XUEPENG
YANG SHIPING
HU BIN
LUO CHEN
WANG CE
TANG XUEYONG

Assignees

贵州电网有限责任公司

Dates

Publication Date: 20260505
Application Date: 20251128

Claims (10)

The MOBO algorithm driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction of key enterprises is characterized by comprising the following steps of, Performing data reading and preprocessing, generating an initial sample, and constructing an objective function based on energy conservation and carbon reduction; Carrying out multi-objective Bayesian optimization initialization on the preprocessed data, and dynamically distributing weights; and generating a candidate decision scheme through iterative optimization by using multi-objective Bayes and carrying out constraint inspection.
2. The method for multi-objective decision making with key enterprise energy saving and carbon reduction driven by MOBO algorithm according to claim 1, wherein the data reading and preprocessing includes reading enterprise time-sharing electricity consumption historical data, real-time electricity price curve, power grid carbon intensity time period data and upper and lower electricity consumption limits of each time period; and cleaning the data, filling the missing value, correcting the abnormal value and ensuring the uniformity of time granularity.
3. The method for key enterprise energy-saving and carbon-reduction multi-objective auxiliary decision-making driven by the MOBO algorithm of claim 2, wherein the generating the initial sample comprises, Generating an initial sample set meeting constraint conditions by taking enterprise time-sharing electricity consumption historical data as a reference; the sample accords with that the total electricity consumption does not exceed the current limit of the enterprise, and the electricity consumption in each period is in the allowed interval.
4. The method for multi-objective decision making with energy conservation and carbon reduction of key enterprises driven by MOBO algorithm according to claim 3, wherein the objective function is constructed based on energy conservation and carbon reduction, which comprises calculating a three-dimensional target value of an initial sample based on real-time data, wherein the total energy consumption is obtained through the electricity consumption of all time periods; Acquiring total carbon emission through the electricity consumption of all time periods and the power grid carbon intensity of the corresponding time period; acquiring total economic cost through the electricity consumption of all time periods and the electricity price of the corresponding time period; And (3) establishing a real-time mapping relation of the three-dimensional objective function, and constructing an energy consumption-carbon row-cost coupling model.
5. The method for performing multi-objective Bayesian optimization initialization of energy conservation and carbon reduction for key enterprises driven by MOBO algorithm according to claim 4, wherein the multi-objective Bayesian optimization initialization includes setting initial sample number, candidate point number of each iteration, maximum iteration number, creating sample matrix X and target value matrix Y, wherein X is used for storing initial samples, Y is used for storing corresponding target values including cost and carbon emission; Using an original curve as a reference, adding random disturbance, generating initial samples meeting the constraint of total power consumption and the constraint of upper and lower limits of power consumption in each period, calculating corresponding cost and carbon emission target values for each initial sample X, and storing the cost and the carbon emission target values in X and Y; constructing a Gaussian process regression model, respectively fitting carbon emission and cost targets by taking time-sharing power consumption as input, respectively aiming at the cost and the carbon emission targets, and training the Gaussian process regression model by using initial sample data And Adopting a constant basis function and ARD Squared Exponential kernel functions, and performing standardization processing; the Gaussian process regression model calculates a predicted value and uncertainty through probabilistic prediction of a kernel function on a target function, and dynamically adjusts the weight of energy consumption, carbon emission and cost targets according to real-time power price fluctuation, carbon quota residual quantity and production load requirements.
6. The method for multi-objective decision making with key enterprise energy saving and carbon reduction driven by MOBO algorithm, as set forth in claim 5, wherein the generating the candidate decision making scheme comprises generating a candidate decision making scheme satisfying time-sharing upper and lower limits and total electric quantity constraint conditions in each iteration, predicting cost and carbon emission target values and corresponding standard deviations of the candidate decision making scheme by using a trained Gaussian process regression model, screening optimal candidate points by expected improvement criteria, and calculating expected improvement values of cost and carbon emission according to the predicted target values and uncertainty: Wherein, the And Respectively the current optimal cost and carbon emission values, And As the predicted target value of the candidate point, And In order to predict the standard deviation of the data, To avoid a minimum value divided by zero, The distribution function is accumulated for a normal distribution, As a function of the normal probability density, And For a standardized value of cost and carbon emissions, And Is a desirable improvement in cost and carbon emissions; Calculating actual target values of the selected candidate points, incorporating a sample set update model, i.e. calculating comprehensive expected improvement values according to dynamic weights : Wherein, the Adjusting the importance of the cost in synthesizing the desired improvement as a weight of the cost; the importance of carbon emissions is adjusted in the overall desired improvement, as a weight for carbon emissions; Selecting a candidate point with the maximum comprehensive expected improvement value as the optimal candidate point and simultaneously as the next evaluation point, calculating the cost and the carbon emission target value of a new selected point, recording the new point and the corresponding target value, and adding the new point to the sample matrix And a target value matrix And retraining a Gaussian process regression model by using the updated data set, and carrying out constraint verification on all candidate schemes, wherein the electricity consumption of each period meets the production requirement in an allowed interval, the total electricity consumption does not exceed the current period allowance of an enterprise, and the total carbon emission does not exceed the carbon allowance.
7. The method for multi-objective decision making with key enterprise energy saving and carbon reduction driven by MOBO algorithm according to claim 6, wherein the iterative optimization includes approximating an optimal solution by updating a sample set and a Gaussian process model, and stopping iteration when the maximum number of iterations or continuous multi-generation target value change is smaller than a set threshold; Based on the iterative optimization result, by comparing the target values of all sample points, a non-dominant solution on the pareto front is identified for each sample point If there is another sample point So that And is also provided with Sample point Instead of the non-dominant solution, extracting the sample and target value corresponding to the non-dominant solution to obtain And Calculating the distances from all pareto solutions to ideal points by using the normalized distances and weights, and selecting the solution with the smallest distance as an optimal solution: Wherein, the As an ideal point of the device, As the worst point of the approach, As a result of the normalization after the result, In order to be a distance from each other, For the weight vector of cost and carbon emissions, I is the variable index for the number of objective functions, For the normalized value of the i-th object, And All target value vectors representing the j-th and i-th sample points; And outputting an optimal time-sharing electricity consumption distribution scheme which comprises the recommended electricity consumption of each time period, and corresponding energy consumption, carbon emission and cost optimization results.
The key enterprise energy-saving and carbon-reduction multi-target auxiliary decision-making system driven by the MOBO algorithm is applied to the key enterprise energy-saving and carbon-reduction multi-target auxiliary decision-making method driven by the MOBO algorithm as claimed in any one of claims 1-7, and is characterized by comprising an objective function construction module, a target function analysis module and a target function analysis module, wherein the objective function construction module is used for reading and preprocessing data to generate an initial sample and constructing an objective function based on energy saving and carbon reduction; The optimizing module is used for carrying out multi-objective Bayesian optimization initialization on the preprocessed data and dynamically distributing weights; and the decision module generates candidate decision schemes through multi-objective Bayesian iterative optimization and performs constraint inspection.
9. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor, when executing the computer program, performs the steps of the MOBO algorithm-driven multi-objective decision-making method for energy conservation and carbon reduction for key enterprises of any one of claims 1 to 7.
10. A computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor performs the steps of the MOBO algorithm driven key enterprise energy saving carbon reduction multi-objective auxiliary decision making method of any one of claims 1 to 7.

Description

MOBO algorithm driven key enterprise energy-saving carbon-reduction multi-target auxiliary decision-making method, system, equipment and medium Technical Field The invention relates to the technical field of industrial energy conservation and emission reduction, in particular to a key enterprise energy conservation and carbon reduction multi-target auxiliary decision-making method, system, equipment and medium driven by MOBO algorithm. Background Worldwide, energy shortage and climate change have become two major challenges threatening sustainable development of human society. In recent years, extreme weather events such as storm flood, high temperature drought and the like frequently occur, and serious impact is caused on an ecological system and the life and economic development of human beings. International Energy Agency (IEA) reports indicate that global energy demand continues to rise over the last few decades, while traditional fossil energy reserves are increasingly diminishing, and stability of energy supply faces tremendous challenges. Meanwhile, the emission of fossil energy causes the aggravation of greenhouse effect and the aggravation of ecological environment, and forms a serious threat to ecological balance and human living environment. Under the large background, energy conservation and carbon reduction become necessary choices for realizing sustainable development in all countries of the world, and are also core tasks for realizing green transformation in various industries, especially in the power industry. Because the power industry is an important area of energy consumption and carbon emission, the carbon emission amount caused by the power industry is more than one fourth of the world energy-related carbon emission amount. Therefore, if the carbon reduction and emission reduction of the electric power industry can be fully realized, the "two-carbon" strategic aim is basically realized. In other words, success or failure of the "two carbon" target is dependent to some extent on the power industry. Currently, low-carbon energy-saving transformation in the power industry is being developed well, on the one hand, clean energy sources such as solar energy, wind energy, water energy, nuclear energy and the like are greatly developed. Taking solar energy as an example, in recent years, the photovoltaic technology is continuously broken through, the photoelectric conversion efficiency of monocrystalline silicon and polycrystalline silicon is continuously improved, the cost is gradually reduced, and the duty ratio of the solar energy in an energy structure is continuously improved. On the other hand, in the energy source conveying and distributing link, the intelligent power grid technology is widely applied. By introducing advanced sensors, communication technology and intelligent control algorithm, the intelligent power grid can realize real-time monitoring and accurate regulation and control of the power system, optimize power distribution, reduce transmission loss and improve energy transmission efficiency. Meanwhile, energy storage technologies such as lithium battery energy storage, pumped storage and the like are utilized to balance energy supply and demand, solve the intermittent problem of renewable energy power generation, and guarantee the stability of energy supply. Under the large background of global energy shortage and climate change, energy conservation and carbon reduction are core tasks of sustainable development of various industries, and the power industry is particularly critical as an important field of energy consumption and carbon emission, and transformation of the power industry is particularly critical. Currently, in the low-carbon energy-saving transformation of the power industry, the development of clean energy, the application of a smart grid, the innovation of a decision model and the like are all progressed, but the decision for enterprise power utilization still has limitations: 1. The multi-focus equipment-level power optimization of the existing energy-saving decision system is not designed aiming at enterprise time-sharing electricity consumption adjustment, and the influence of energy prices and carbon intensity differences in different time periods on multi-objective coordination is ignored; 2. The static decision model cannot respond to dynamic factors such as time-of-use electricity price fluctuation, power grid carbon intensity time period change and the like; 3. tools for quantitatively evaluating the multidimensional influence of different time-sharing electricity utilization decisions on energy consumption, carbon emission and economic cost are lacking. Therefore, an innovative multi-target energy-saving carbon reduction auxiliary decision making system and method are needed, the time-sharing electricity consumption of a focused enterprise is adjusted, multi-target factors are comprehensively considered, and the decision scientificity is improved. Di