Energy-saving and carbon-reducing multi-target auxiliary decision-making method, system and equipment for important enterprise driven by MOBO algorithm, and medium

Abstract

The invention discloses an MOBO algorithm-driven key enterprise energy-saving and carbon-reducing multi-objective auxiliary decision-making method, system, equipment and medium, and belongs to the technical field of industrial energy saving and emission reduction, and the method comprises the steps: carrying out the data reading and preprocessing, generating an initial sample, and constructing an objective function based on energy saving and carbon reduction; performing multi-target Bayesian optimization initialization on the preprocessed data, and dynamically allocating weights; and performing iterative optimization through multi-objective Bayesian to generate a candidate decision scheme, and performing constraint check. The method responds to the time-of-use electricity price, the carbon factor and the load state, and the problem of time-of-use electricity multi-target separation is solved; an MOBO algorithm is adopted, a target function is modeled through a Gaussian process regression model, external condition mutation can be quickly responded, complex constraints can be effectively processed, the performance of multi-target optimization in a dynamic energy-saving and carbon-reducing scene is remarkably improved, the convergence speed is higher, and the optimization effect is better.

Description

Technical Field

[0001] This invention relates to the field of industrial energy conservation and emission reduction technology, specifically to a multi-objective auxiliary decision-making method, system, equipment, and medium for energy conservation and carbon reduction in key enterprises driven by the MOBO algorithm. Background Technology

[0002] Globally, energy shortages and climate change have become two major challenges threatening the sustainable development of human society. In recent years, extreme weather events such as torrential rains and floods, as well as extreme heat and droughts, have occurred frequently, causing severe impacts on ecosystems, human lives, and economic development. According to a report by the International Energy Agency (IEA), global energy demand has continued to rise over the past few decades, while traditional fossil fuel reserves have been dwindling, posing a significant challenge to the stability of energy supply. At the same time, emissions from fossil fuels have exacerbated the greenhouse effect and degraded the ecological environment, posing a serious threat to ecological balance and the human living environment.

[0003] Against this backdrop, energy conservation and carbon reduction have become an inevitable choice for countries worldwide to achieve sustainable development, and a core task for various industries, especially the power sector, to achieve green transformation. This is because the power sector is a major contributor to energy consumption and carbon emissions, accounting for more than a quarter of global energy-related carbon emissions. Therefore, if carbon reduction and emission reduction in the power sector can be fully implemented, the "dual-carbon" strategic goal will be essentially achieved. In other words, the success or failure of the "dual-carbon" goal depends to a certain extent on the power sector.

[0004] Currently, the power industry's low-carbon and energy-saving transformation is in full swing. On the one hand, it is vigorously developing clean energy sources such as solar, wind, hydro, and nuclear power. Taking solar energy as an example, photovoltaic technology has made continuous breakthroughs in recent years, with the photoelectric conversion efficiency of monocrystalline and polycrystalline silicon continuously improving and costs gradually decreasing, leading to a continuous increase in the proportion of solar energy in the energy structure. On the other hand, smart grid technology is widely used in energy transmission and distribution. By introducing advanced sensors, communication technologies, and intelligent control algorithms, smart grids can achieve real-time monitoring and precise regulation of the power system, optimize power distribution, reduce transmission losses, and improve energy transmission efficiency. Simultaneously, energy storage technologies, such as lithium battery energy storage and pumped hydro storage, are used to balance energy supply and demand, solve the intermittency problem of renewable energy power generation, and ensure the stability of energy supply.

[0005] Against the backdrop of global energy shortages and climate change, energy conservation and carbon reduction are core tasks for sustainable development across all industries. The power industry, as a major consumer of energy and emitter of carbon, faces particularly critical challenges in its transformation. Currently, progress has been made in the power industry's low-carbon and energy-saving transformation, including the development of clean energy, the application of smart grids, and innovation in decision-making models. However, limitations still exist in areas such as enterprise electricity consumption decisions. 1. Existing energy-saving decision-making systems mostly focus on equipment-level power optimization, without designing for time-of-use electricity consumption adjustment for enterprises, and ignore the impact of energy price and carbon intensity differences at different times on multi-objective coordination; 2. Static decision-making models cannot respond to dynamic factors such as time-of-use electricity price fluctuations and inter-period changes in grid carbon intensity; 3. There is a lack of tools to quantitatively assess the multidimensional impact of different time-of-use electricity decisions on energy consumption, carbon emissions, and economic costs.

[0006] Therefore, an innovative multi-objective energy-saving and carbon-reduction auxiliary decision-making system and method are needed, which focuses on the time-of-use electricity consumption adjustment of enterprises, comprehensively considers multiple objective factors, and improves the scientific nature of decision-making. Summary of the Invention

[0007] In view of the above-mentioned problems, the present invention is proposed.

[0008] Therefore, the key problem that this invention aims to solve is that current energy-saving decision-making systems have significant limitations, specifically: they often only optimize energy consumption or carbon emissions in isolation, ignoring the synergy between multiple objectives; they use static decision-making models, which cannot respond to dynamic factors such as real-time energy price fluctuations; and they lack effective tools to quantify and assess the multidimensional impact of different decisions on "energy consumption-carbon emissions-economic costs".

[0009] To address the aforementioned technical problems, this invention provides the following technical solution: a multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises driven by the MOBO algorithm, comprising, Data is read and preprocessed to generate initial samples, and an objective function is constructed based on energy conservation and carbon reduction. The preprocessed data is initialized using multi-objective Bayesian optimization and weights are dynamically allocated. Candidate decision schemes are generated through iterative optimization using multi-objective Bayesian optimization and constraint verification is performed.

[0010] As a preferred embodiment of the MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises as described in this invention, the data reading and preprocessing includes reading the enterprise's time-of-use electricity consumption history data, real-time electricity price curve, grid carbon intensity time period data, and the upper and lower limits of electricity consumption for each time period. The data is cleaned, missing values ​​are filled in, outliers are corrected, and time granularity is ensured to be consistent.

[0011] As a preferred embodiment of the MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises as described in this invention, the generation of initial samples includes: Based on the historical data of enterprises' time-of-use electricity consumption, an initial sample set that meets the constraints is generated; The sample meets the requirement that the total electricity consumption does not exceed the enterprise's current quota, and the electricity consumption in each time period is within the allowable range.

[0012] As a preferred embodiment of the MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises described in this invention, the objective function based on energy conservation and carbon reduction includes: calculating the three-dimensional objective value of the initial sample based on real-time data: obtaining the total energy consumption through the electricity consumption of all time periods; Total carbon emissions are obtained by comparing electricity consumption over all time periods with the corresponding grid carbon intensity. The total economic cost is obtained by comparing electricity consumption over all time periods with the corresponding electricity prices for those periods. Establish a real-time mapping relationship for the three-dimensional objective function and construct an energy consumption-carbon emission-cost coupling model.

[0013] The preferred technical solution in this embodiment of the invention has the following advantages: by cleaning the data and generating the initial samples, the accuracy and consistency of the model input data are ensured, laying a reliable foundation for subsequent optimization.

[0014] As a preferred embodiment of the MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises described in this invention, the multi-objective Bayesian optimization initialization includes setting the initial number of samples, the number of candidate points in each iteration, the maximum number of iterations, and creating a sample matrix X and an objective value matrix Y, wherein X is used to store the initial samples, and Y is used to store the corresponding objective values, including cost and carbon emissions. Using the original curve as a baseline, random perturbations are added to generate initial samples that satisfy the total electricity consumption constraint and the upper and lower limits of electricity consumption in each time period. For each initial sample x, the corresponding cost and carbon emission target values ​​are calculated and stored in X and Y. A Gaussian process regression model is constructed, using time-of-use electricity consumption as input, to fit carbon emission and cost targets respectively; Gaussian process regression models are trained using initial sample data for cost and carbon emission targets respectively. and Constant basis functions and ARD Squared Exponential kernel functions are used, and standardization is performed. The Gaussian process regression model uses a kernel function to probabilistically predict the objective function, calculates the predicted value and uncertainty, and dynamically adjusts the weights of energy consumption, carbon emissions, and cost targets based on real-time electricity price fluctuations, remaining carbon quotas, and production load demand.

[0015] The beneficial effects of the preferred technical solution in the embodiments of the present invention are as follows: by using a Gaussian process model and dynamic weight allocation, the efficiency and adaptability of the optimization process are improved, and energy consumption, carbon emissions and cost targets are effectively balanced.

[0016] As a preferred embodiment of the MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises as described in this invention, the generation of candidate decision schemes includes: in each iteration, generating candidate decision schemes that satisfy the time-sharing upper and lower limits and total power consumption constraints; using a trained Gaussian process regression model to predict the cost and carbon emission target values ​​and corresponding standard deviations of the candidate decision schemes; selecting the optimal candidate point through the expected improvement criterion; and calculating the expected improvement values ​​of cost and carbon emissions based on the predicted target values ​​and uncertainties. in, and These represent the current optimal cost and carbon emission values, respectively. and The target value for the candidate point is... and The standard deviation of the forecast. To avoid the minimum value when divided by zero, The cumulative distribution function is the normal distribution. It is a normal probability density function. and Standardized values ​​for cost and carbon emissions. and The expected improvement in cost and carbon emissions; The actual target value of the selected candidate point is calculated and incorporated into the sample set to update the model; that is, the comprehensive expected improvement value is calculated based on the dynamic weights. : in, Assigning a weight to cost, and adjusting the importance of cost in the overall expected improvement; The weighting of carbon emissions is adjusted to reflect the overall expected improvement in carbon emissions; The candidate point with the largest comprehensive expected improvement value is selected as the best candidate point and also serves as the next evaluation point. The cost and carbon emission target value of the new selected point are calculated, the new point and its corresponding target value are recorded, and the new point is added to the sample matrix. and target value matrix In this process, the Gaussian process regression model is retrained using the updated dataset; constraints are checked on all candidate schemes: electricity consumption in each time period meets production needs within the allowable range; total electricity consumption does not exceed the enterprise's current quota, and total carbon emissions do not exceed the carbon quota.

[0017] As a preferred embodiment of the MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises described in this invention, the iterative optimization includes approximating the optimal solution by updating the sample set and the Gaussian process model, and stopping the iteration when the maximum number of iterations is reached or the change of the target value for multiple consecutive generations is less than a set threshold. Based on the iterative optimization results, non-dominated solutions on the Pareto front are identified by comparing the objective values ​​of all sample points. For each sample point... If another sample point exists , making and Then the sample points If the solution is not non-dominated, extract the samples and target values ​​corresponding to the non-dominated solutions to obtain... and Using normalized distance and weights, calculate the distance from all Pareto solutions to the ideal point, and select the solution with the smallest distance as the optimal solution. in, For the ideal point, As the worst point, For the standardized results, For distance, As a weighted vector of cost and carbon emissions, Let i be the number of objective functions, and i be the variable index. Let be the standardized value of the i-th objective. and Let represent the vector of all target values ​​for the j-th and i-th sample points; Output the optimal time-of-use electricity allocation scheme, including the suggested electricity consumption for each time period, and the corresponding energy consumption, carbon emissions, and cost optimization results.

[0018] The beneficial effects of the preferred technical solution in the embodiments of the present invention are as follows: by using the expected improvement criteria and constraint verification, the Pareto optimal solution is searched efficiently, and a feasible and optimized energy-saving and carbon-reducing power allocation scheme is output.

[0019] Another objective of this invention is to provide a multi-objective auxiliary decision-making system for energy conservation and carbon reduction in key enterprises, driven by the MOBO algorithm.

[0020] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a multi-objective auxiliary decision-making system for energy conservation and carbon reduction of key enterprises driven by MOBO algorithm, comprising: an objective function construction module, which performs data reading and preprocessing, generates initial samples, and constructs an objective function based on energy conservation and carbon reduction; The optimization module performs multi-objective Bayesian optimization initialization on the preprocessed data and dynamically allocates weights; The decision-making module generates candidate decision schemes through multi-objective Bayesian iterative optimization and performs constraint checks.

[0021] The present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises.

[0022] The present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises.

[0023] The beneficial effects of this invention are as follows: This invention proposes a dynamic multi-objective collaborative optimization mechanism based on the MOBO algorithm. It quantifies the two-dimensional objective function of carbon emissions and economic costs in real time while maintaining the basic electricity consumption required for enterprise production. Combined with a dynamic weight allocation strategy, it responds to time-of-use electricity prices, carbon factors and load status, thus solving the problem of fragmented multi-objectives in time-of-use electricity consumption.

[0024] The MOBO algorithm is adopted, and the objective function is modeled by a Gaussian process regression model. Combined with the expected improvement criterion to balance the exploration and utilization relationship, it can quickly respond to sudden changes in external conditions and effectively handle complex constraints. It significantly improves the performance of multi-objective optimization in dynamic energy-saving and carbon-reduction scenarios, with faster convergence speed and better optimization effect. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 The above is a flowchart of a multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises driven by the MOBO algorithm, provided in one embodiment of the present invention.

[0027] Figure 2 The graph shows the time-sharing factor variation of the data running the MOBO algorithm optimization results of a key enterprise energy conservation and carbon reduction multi-objective auxiliary decision-making method provided in an embodiment of the present invention.

[0028] Figure 3The Pareto front plot of the data running MOBO algorithm optimization results of the key enterprise energy conservation and carbon reduction multi-objective auxiliary decision-making method provided in an embodiment of the present invention.

[0029] Figure 4 This is a comparison chart of electricity consumption before and after optimization, based on the MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises, provided in one embodiment of the present invention. Detailed Implementation

[0030] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0031] Example 1, referring to Figure 1 This is one embodiment of the present invention, which provides a MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises, including: S100: Perform data reading and preprocessing to generate initial samples, and construct an objective function based on energy conservation and carbon reduction; S200. Perform multi-objective Bayesian optimization initialization on the preprocessed data and dynamically allocate weights; S300: Iterative optimization using multi-objective Bayesian methods generates candidate decision schemes and performs constraint checks. It should be noted that existing systems generally adopt a single-objective optimization model at the equipment level, without establishing a collaborative optimization mechanism between time-of-use electricity consumption and energy consumption, carbon emissions, and economic benefits. This leads to an inherent conflict between time-of-use electricity measures and carbon reduction and cost objectives, making it impossible to achieve overall system optimization. Traditional decision-making models use static parameter systems, which cannot dynamically respond to real-time variables such as fluctuations in time-of-use electricity prices and changes in grid carbon intensity over time. Model updates lag behind actual operating conditions, resulting in decision-making biases. Existing technologies lack comprehensive quantitative assessment tools for the two-dimensional targets of carbon emissions and operating costs associated with time-of-use electricity consumption, making it difficult to accurately analyze the comprehensive performance of different time-of-use decision-making schemes and restricting scientific decision-making capabilities.

[0032] Therefore, addressing the aforementioned problems, this invention, through steps S100-S300, systematically solves the limitations of existing energy-saving decision-making systems by constructing a multi-objective optimization mechanism and decision path optimization engine based on the MOBO algorithm. Specifically, firstly, a real-time quantitative model of energy intensity, carbon emissions, and operating costs is established to accurately characterize multi-dimensional objectives; simultaneously, a dynamic weight allocation strategy is designed to dynamically adjust the weights of each objective based on real-time electricity price signals and dynamic changes in carbon factors, ensuring the flexibility and adaptability of the decision-making process; finally, the MOBO algorithm is used to construct a decision path optimization engine, effectively handling high-dimensional non-convex optimization problems, achieving collaborative optimization and dynamic balance of multiple objectives, and providing comprehensive support for energy-saving and carbon-reduction decisions in complex scenarios.

[0033] Example 2, refer to Figure 1 This is one embodiment of the present invention, which provides a MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises, including: In this embodiment of the invention, data reading and preprocessing are performed in S100 to generate an initial sample, and an objective function is constructed based on energy conservation and carbon reduction, including the following steps S101-S103: S101. Read the enterprise's historical time-of-use electricity consumption data, real-time electricity price curve, grid carbon intensity time period data, and the upper and lower limits of electricity consumption for each time period; Specifically, relevant data is read, including time-of-use electricity consumption limits, actual electricity consumption data, electricity price data, carbon factor data, etc. The data is preprocessed to ensure its validity and integrity and to prevent data loss or anomalies. Read the upper and lower limits of power consumption (24-hour data) and obtain the minimum value. and maximum value ; Read actual power consumption data (24 hours) to obtain raw power consumption; Read electricity price data (24 hours) to obtain the electricity price; Read carbon factor data (24 hours) to obtain carbon factor; Ensure all data is 24-hour data, and convert all vectors into column vectors to calculate the original total electricity consumption, original cost, and original carbon emissions; In an embodiment of the present invention, the preprocessing includes the following step A1: A1. Clean the data, fill in missing values, correct outliers, and ensure consistent time granularity.

[0034] In an optional implementation, the preprocessing in S102 can check for missing values ​​in the data, and for any missing time period, use linear interpolation to fill in the missing data based on the data values ​​of the adjacent time periods; set a fixed threshold and replace data points exceeding the threshold with the historical average of that time period; force all data to be aligned to a 24-hour time point, and if the data time granularity is inconsistent, use nearest neighbor interpolation to fill in the missing time period.

[0035] In another alternative implementation, the preprocessing in S102 can also be based on a rule-based preprocessing method. For missing values, data from the same day of the previous week is used to fill in the missing values. If there is no historical data, the overall average value is used to fill in the missing values. Based on business rules (such as electricity consumption should not exceed the theoretical lower limit), outliers are directly truncated to the allowable range. The data is aggregated into hourly levels through a simple resampling method. If data points are missing, the values ​​from the previous period are directly copied.

[0036] In an embodiment of the present invention, S102, energy consumption-carbon emission-cost coupling modeling, includes the following steps B1-B2: B1. Calculate the three-dimensional target value of the initial sample based on real-time data: Total energy consumption is the sum of electricity consumption in each time period; Total carbon emissions are the sum of the products of electricity consumption in each time period and the corresponding carbon intensity of the power grid in that time period; The total economic cost is the sum of the products of electricity consumption in each time period and the corresponding electricity price in that time period.

[0037] B2. Establish the real-time mapping relationship of the three-dimensional objective function: in, For energy intensity model, Electricity consumption during period t, unit: kWh; For carbon emission models, Let t be the carbon intensity of the power grid. This is an economic cost model; T represents the total number of time periods (e.g., 24 time periods per day). The time interval is set to 1 hour, depending on the desired level of precision. Electricity costs are updated in real time.

[0038] In an optional implementation, the objective function constructed in S102 can be modeled based on the three-dimensional objective function of the time period average, and the three-dimensional objective values ​​of the initial sample are calculated: average energy consumption, average carbon emission, and average economic cost. The real-time mapping relationship of the three-dimensional objective function is established, and the objective functions are average energy consumption, average carbon emission, and average economic cost, respectively. However, the calculation of the average value smooths out the outliers during peak periods, resulting in insufficient optimization effect for key periods (such as high energy consumption or high carbon emission periods).

[0039] In another optional implementation, the objective function in S102 can also be modeled using a three-dimensional objective function with fixed weighting factors. Fixed weighting factors are preset for each time period, and the three-dimensional objective values ​​of the initial sample are calculated: total energy consumption, total carbon emissions, and total economic cost. A real-time mapping relationship of the three-dimensional objective function is established, with the objective functions being weighted total energy consumption, weighted total carbon emissions, and weighted total economic cost, respectively. However, the fixed weighting factors cannot dynamically adapt to real-time changing market conditions, causing the optimization results to deviate from the actual optimal solution.

[0040] In an embodiment of the present invention, S103, generating an initial sample, includes the following steps C1-C2: C1. Set the initial number of samples, the number of candidate points in each iteration, and the maximum number of iterations. Create a sample matrix X and a target value matrix Y, where X is used to store the initial samples and Y is used to store the corresponding target values, including cost and carbon emissions. C2. Using the original curve as a baseline, add random perturbations to generate initial samples that satisfy the total electricity consumption constraint and the upper and lower limits of electricity consumption in each time period. For each initial sample x, calculate the corresponding cost and carbon emission target values ​​and store them in X and Y.

[0041] In an optional implementation, the initial sample generated in S103 can be generated based on the uniformly sampled initial sample. Using the original electricity consumption curve as a reference, the upper and lower limits of electricity consumption for each time period are obtained. For each initial sample, a value is randomly sampled uniformly from the upper and lower limits of electricity consumption for each time period independently to form a new electricity consumption curve. It is checked whether the total electricity consumption of the generated curve is close to the original total electricity consumption (allowing a certain deviation, such as ±5%). If it does not meet the requirements, the values ​​of some time periods are scaled or adjusted to meet the overall constraints. A specified number of initial samples are generated, and the cost and economic carbon emission target values ​​corresponding to each sample are calculated and stored in the sample matrix X and the target value matrix Y.

[0042] In another optional implementation, the initial sample generation in S103 can also be based on Gaussian perturbation. Using the original electricity consumption curve as a benchmark, the average electricity consumption for each time period is calculated, or the original value is directly used as the center. Gaussian noise (normal distribution perturbation) is added to the electricity consumption for each time period. The mean of the noise is set to 0, and the standard deviation is set according to the upper and lower limits of electricity consumption (e.g., 10% of the difference between the upper and lower limits) to ensure that the perturbed electricity consumption does not exceed the upper and lower limits of each time period. If it does, it is truncated to the boundary value. The total electricity consumption constraint is checked. If the deviation from the original total electricity consumption is too large, the total electricity consumption is brought back to the allowable range through linear adjustment. A specified number of initial samples are generated, and the cost and economic carbon emission target values ​​corresponding to each sample are calculated and stored in the sample matrix X and the target value matrix Y.

[0043] In this embodiment of the invention, step S200 involves initializing the preprocessed data using multi-objective Bayesian optimization and dynamically assigning weights, including the following steps S201-S202: In an embodiment of the present invention, S201, optimization initialization, includes the following steps D1-D2: D1. Construct a Gaussian process regression model, using time-of-use electricity consumption as input, and fit it to carbon emission and cost targets respectively; D2. Train Gaussian process regression models using initial sample data, targeting cost and carbon emission goals respectively. and It employs constant basis functions and the ARD Squared Exponential kernel function, and performs standardization processing.

[0044] In an optional implementation, the optimization initialization in S201 can be performed using a random forest regression model. The random forest regression model is constructed with time-of-use electricity consumption as input and is fitted to carbon emission and cost targets respectively. The random forest regression model is trained using initial sample data for cost and carbon emission targets respectively, and the data is standardized. However, it is prone to overfitting and has low accuracy in estimating uncertainty when dealing with small sample data.

[0045] In another alternative implementation, the optimization initialization in S201 can also be performed using a support vector regression model. The support vector regression model is constructed with time-of-use electricity consumption as input, and is fitted to carbon emission and cost targets respectively. The support vector regression model is trained using initial sample data for cost and carbon emission targets respectively, and the data is standardized. However, it is sensitive to the selection of hyperparameters, and the training time increases significantly with the increase of data volume.

[0046] S202, the Gaussian process regression model uses a kernel function to probabilistically predict the objective function, calculate the predicted value and uncertainty; and dynamically adjusts the weights of energy consumption, carbon emissions, and cost targets based on real-time electricity price fluctuations, remaining carbon quotas, and production load demand. Dynamic weights (cost weights gradually increase with iterations): in, This represents the current iteration number. The maximum number of iterations, Assigning a weight to cost, and adjusting the importance of cost in the overall expected improvement; The weighting of carbon emissions is adjusted to reflect the overall expected improvement in carbon emissions; In an embodiment of the present invention, step S300 involves iterative optimization using multi-objective Bayesian methods to generate candidate decision schemes and perform constraint checks, including the following steps S301-S302: S301. In each iteration, candidate decision schemes that satisfy the time-sharing upper and lower limits and total energy consumption constraints are generated. Using a trained Gaussian process regression model, the cost and carbon emission target values ​​and corresponding standard deviations of the candidate decision schemes are predicted. The optimal candidate point is selected through the expected improvement criterion. Based on the predicted target values ​​and uncertainties, the expected improvement values ​​of cost and carbon emissions are calculated. in, and These represent the current optimal cost and carbon emission values, respectively. and The target value for the candidate point is... and The standard deviation of the forecast. To avoid the minimum value when divided by zero, The cumulative distribution function is the normal distribution. It is a normal probability density function. and Standardized values ​​for cost and carbon emissions. and The expected improvement in cost and carbon emissions; The actual target value of the selected candidate point is calculated and incorporated into the sample set to update the model; that is, the comprehensive expected improvement value is calculated based on the dynamic weights. : The candidate point with the largest comprehensive expected improvement value is selected as the best candidate point and also serves as the next evaluation point. The cost and carbon emission target value of the new selected point are calculated, the new point and its corresponding target value are recorded, and the new point is added to the sample matrix. and target value matrix In the process, the Gaussian process regression model is retrained using the updated dataset; Constraint verification was performed on all candidate solutions: electricity consumption in each time period was within the allowable range to meet production needs; total electricity consumption did not exceed the enterprise's current quota, and total carbon emissions did not exceed the carbon allowance. Specifically, during the generation of initial samples and candidate points, ensure that the following constraints are met: in, This means that the sum of electricity consumption in all time periods equals the original total electricity consumption, keeping the total energy consumption constant; This indicates that the electricity consumption for each time period must be within the minimum and maximum allowable range for that time period, taking into account both production needs and time period control requirements.

[0047] S302. By updating the sample set and Gaussian process model, the optimal solution is approximated. The iteration stops when the maximum number of iterations is reached or the change of the target value is less than the set threshold for multiple consecutive generations. Based on the iterative optimization results, non-dominated solutions on the Pareto front are identified by comparing the objective values ​​of all sample points. For each sample point... If another sample point exists , making and Then the sample points If the solution is not non-dominated, extract the samples and target values ​​corresponding to the non-dominated solutions to obtain... and Using normalized distance and weights, calculate the distance from all Pareto solutions to the ideal point, and select the solution with the smallest distance as the optimal solution. in, For the ideal point, As the worst point, For the standardized results, For distance, As a weighted vector of cost and carbon emissions, this is used here. Cost weight carbon emission weight ; Let i be the number of objective functions, and i be the variable index. Let be the standardized value of the i-th objective. and Let represent the vector of all target values ​​for the j-th and i-th sample points; The first objective is to achieve the ideal cost level. The second objective is the ideal point for carbon emissions; similarly, and The worst point; Output the optimal time-of-use electricity allocation scheme, including the suggested electricity consumption for each time period, and the corresponding energy consumption, carbon emissions, and cost optimization results.

[0048] Example 3, referring to Figures 2-4This invention provides a MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises. To verify the beneficial effects of this invention, scientific demonstration is carried out through experiments.

[0049] Regarding the decision-making scheme selection in the embodiment, data were collected from multiple high-energy-consuming enterprises for comparison before and after optimization, and the results are shown in Table 1. Table 1. Overview of optimization results for multiple high-energy-consuming enterprises in a certain city in a certain year and month.

[0050] Through the above process, the MOBO algorithm utilizes probabilistic modeling and efficient search capabilities to quickly find the optimal solution for multi-objective balance under dynamic electricity price and carbon intensity scenarios, achieving synergistic optimization of carbon emission and cost factors under the premise of meeting basic production energy consumption requirements.

[0051] Example 4 is an embodiment of the present invention. The above is an illustrative scheme of the MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises. It should be noted that the technical solution of the MOBO algorithm-driven multi-objective auxiliary decision-making system for energy conservation and carbon reduction in key enterprises belongs to the same concept as the technical solution of the MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises described above. For details not described in detail in the technical solution of the MOBO algorithm-driven multi-objective auxiliary decision-making system for energy conservation and carbon reduction in key enterprises in this embodiment, please refer to the description of the technical solution of the MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises described above.

[0052] This embodiment provides a MOBO algorithm-driven multi-objective auxiliary decision-making system for energy conservation and carbon reduction in key enterprises, including: an objective function construction module, which performs data reading and preprocessing, generates initial samples, and constructs an objective function based on energy conservation and carbon reduction; The optimization module performs multi-objective Bayesian optimization initialization on the preprocessed data and dynamically allocates weights; The decision-making module generates candidate decision schemes through multi-objective Bayesian iterative optimization and performs constraint checks.

[0053] This embodiment also provides an electronic device applicable to the MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises as proposed in the above embodiment.

[0054] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises as proposed in the above embodiment.

[0055] The storage medium proposed in this embodiment and the multi-objective auxiliary decision-making method for energy saving and carbon reduction of key enterprises driven by MOBO algorithm proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0056] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0057] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises driven by the MOBO algorithm, characterized by: include, Data reading and preprocessing are performed to generate initial samples, and an objective function is constructed based on energy conservation and carbon reduction. The preprocessed data is initialized using multi-objective Bayesian optimization, and weights are dynamically allocated. Candidate decision schemes are generated through iterative optimization using multi-objective Bayesian methods, and constraint checks are performed.

2. The MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises as described in claim 1, characterized in that: The data reading and preprocessing includes reading the enterprise's historical time-of-use electricity consumption data, real-time electricity price curves, grid carbon intensity data for different time periods, and the upper and lower limits of electricity consumption for each time period. Data is cleaned, missing values ​​are filled in, outliers are corrected, and time granularity is ensured to be consistent.

3. The MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises as described in claim 2, characterized in that: The generation of the initial sample includes, Based on the historical data of enterprises' time-of-use electricity consumption, an initial sample set that meets the constraints is generated; The sample meets the requirement that the total electricity consumption does not exceed the enterprise's current quota, and the electricity consumption in each time period is within the allowable range.

4. The MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises as described in claim 3, characterized in that: The objective function based on energy conservation and carbon reduction includes calculating the three-dimensional objective value of the initial sample based on real-time data: obtaining the total energy consumption through electricity consumption in all time periods; Total carbon emissions are obtained by comparing electricity consumption over all time periods with the corresponding grid carbon intensity. The total economic cost is obtained by comparing electricity consumption over all time periods with the corresponding electricity prices for those periods. Establish a real-time mapping relationship for the three-dimensional objective function and construct an energy consumption-carbon emission-cost coupling model.

5. The MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises as described in claim 4, characterized in that: The initialization of multi-objective Bayesian optimization includes setting the initial number of samples, the number of candidate points in each iteration, the maximum number of iterations, and creating a sample matrix X and an objective value matrix Y, where X is used to store the initial samples and Y is used to store the corresponding objective values, including cost and carbon emissions. Using the original curve as a baseline, random perturbations are added to generate initial samples that satisfy the total electricity consumption constraint and the upper and lower limits of electricity consumption in each time period. For each initial sample x, the corresponding cost and carbon emission target values ​​are calculated and stored in X and Y. A Gaussian process regression model is constructed, using time-of-use electricity consumption as input, to fit carbon emission and cost targets respectively; Gaussian process regression models are trained using initial sample data for cost and carbon emission targets respectively. and Constant basis functions and ARD Squared Exponential kernel functions are used, and standardization is performed. The Gaussian process regression model uses a kernel function to probabilistically predict the objective function, calculates the predicted value and uncertainty, and dynamically adjusts the weights of energy consumption, carbon emissions, and cost targets based on real-time electricity price fluctuations, remaining carbon quotas, and production load demand.

6. The MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises as described in claim 5, characterized in that: The generation of candidate decision schemes includes, in each iteration, generating candidate decision schemes that satisfy the time-sharing upper and lower limits and the total energy consumption constraints; using a trained Gaussian process regression model to predict the cost and carbon emission target values ​​and corresponding standard deviations of the candidate decision schemes; selecting the optimal candidate point through the expected improvement criterion; and calculating the expected improvement values ​​of cost and carbon emissions based on the predicted target values ​​and uncertainties. in, and These represent the current optimal cost and carbon emission values, respectively. and The target value for the candidate point is... and The standard deviation of the forecast. To avoid the minimum value when divided by zero, The cumulative distribution function is the normal distribution. It is a normal probability density function. and Standardized values ​​for cost and carbon emissions. and The expected improvement in cost and carbon emissions; The actual target value of the selected candidate point is calculated and incorporated into the sample set to update the model; that is, the comprehensive expected improvement value is calculated based on the dynamic weights. : in, Assigning a weight to cost, and adjusting the importance of cost in the overall expected improvement; The weighting of carbon emissions is adjusted to reflect the overall expected improvement in carbon emissions; The candidate point with the largest comprehensive expected improvement value is selected as the best candidate point and also serves as the next evaluation point. The cost and carbon emission target value of the new selected point are calculated, the new point and its corresponding target value are recorded, and the new point is added to the sample matrix. and target value matrix In this process, the Gaussian process regression model is retrained using the updated dataset; constraints are checked on all candidate schemes: electricity consumption in each time period meets production needs within the allowable range; total electricity consumption does not exceed the enterprise's current quota, and total carbon emissions do not exceed the carbon quota.

7. The MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises as described in claim 6, characterized in that: The iterative optimization includes approximating the optimal solution by updating the sample set and the Gaussian process model, and stopping the iteration when the maximum number of iterations is reached or the change of the target value is less than a set threshold for multiple consecutive generations. Based on the iterative optimization results, non-dominated solutions on the Pareto front are identified by comparing the objective values ​​of all sample points. For each sample point... If another sample point exists , making and Then the sample points If the solution is not non-dominated, extract the samples and target values ​​corresponding to the non-dominated solutions to obtain... and Using normalized distance and weights, calculate the distance from all Pareto solutions to the ideal point, and select the solution with the smallest distance as the optimal solution. in, For the ideal point, As the worst point, For the standardized results, For distance, As a weighted vector of cost and carbon emissions, Let i be the number of objective functions, and i be the variable index. Let be the standardized value of the i-th objective. and Let represent the vector of all target values ​​for the j-th and i-th sample points; Output the optimal time-of-use electricity allocation scheme, including the suggested electricity consumption for each time period, and the corresponding energy consumption, carbon emissions, and cost optimization results.

8. A MOBO algorithm-driven multi-objective auxiliary decision-making system for energy conservation and carbon reduction in key enterprises, employing the MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises as described in any one of claims 1 to 7, characterized in that... include: The objective function construction module reads and preprocesses data, generates initial samples, and constructs an objective function based on energy conservation and carbon reduction. The optimization module performs multi-objective Bayesian optimization initialization on the preprocessed data and dynamically allocates weights; The decision-making module generates candidate decision schemes through multi-objective Bayesian iterative optimization and performs constraint checks.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the MOBO algorithm-driven multi-objective auxiliary decision-making method for energy conservation and carbon reduction in key enterprises as described in any one of claims 1 to 7.

Images