Comprehensive energy system optimization scheduling model building and solving method considering uncertainty carbon transaction

Abstract

The invention discloses a comprehensive energy system optimization scheduling model construction method considering uncertainty carbon transactions, comprising the following steps: predicting an obtained historical carbon price sequence to obtain a predicted carbon price; according to the predicted carbon price, establishing an uncertainty carbon transaction model based on carbon price prediction; establishing an improved conditional value-at-risk model based on the uncertain carbon transaction according to the carbon price prediction-based uncertain carbon transaction model; and according to the carbon price prediction-based uncertain carbon transaction model and the improved conditional value-at-risk model based on the uncertain carbon transaction, establishing a comprehensive energy system optimization scheduling model considering the uncertain carbon transaction. Furthermore, a quadratic programming hybrid algorithm based on a cross genetic mechanism is adopted to solve the integrated energy system optimal scheduling model considering the uncertain carbon transactions. The method has high superiority in the aspects of carbon capture level and total operation cost of the comprehensive energy system.

Description

TECHNICAL FIELD

[0001] The application relates to a method for constructing and solving an optimization scheduling model of a comprehensive energy system considering uncertain carbon trading, and belongs to the field of comprehensive energy system scheduling. BACKGROUND

[0002] With the tightening of fossil resources and the aggravation of environmental problems, carbon emission reduction has become an urgent challenge faced by all countries. The carbon trading mechanism can optimize the allocation of system resources and promote energy saving and emission reduction, and is therefore widely used in comprehensive energy systems.

[0003] The carbon price is affected by many factors such as carbon market policy, market supply and demand status, and total carbon emission, and presents the characteristics of real-time fluctuation. This uncertainty makes it difficult to predict the carbon price. If the uncertainty of the carbon price is not fully considered in the optimization scheduling process of the comprehensive energy system, the scheduling scheme of the comprehensive energy system may face high market risks due to being too conservative or aggressive. In view of this, in order to effectively reduce the risk of the comprehensive energy system participating in carbon market trading, constructing an accurate carbon price prediction mechanism has become one of the key paths to realize the low-carbon economic optimization scheduling of the comprehensive energy system. The optimization scheduling of the comprehensive energy system considering uncertain carbon trading and risk assessment can coordinate multi-energy complementation and uncertain carbon trading, improve the economic benefit of the system, reduce the carbon emission of the system, and provide a reference for decision makers participating in carbon trading by predicting the carbon price.

[0004] Therefore, the application is proposed. SUMMARY

[0005] The application provides a method for constructing an optimization scheduling model of a comprehensive energy system considering uncertain carbon trading, which is used to construct an optimization scheduling model of a comprehensive energy system considering uncertain carbon trading with the comprehensive energy system operation cost, the comprehensive energy system purchase cost of energy from a superior energy, the carbon trading cost of the comprehensive energy system, the CO2 storage and transportation cost generated by the comprehensive energy system, and the conditional risk cost of carbon trading as an objective function. Further, a quadratic programming hybrid algorithm based on a cross genetic mechanism is used to solve the optimization scheduling model of the comprehensive energy system considering uncertain carbon trading.

[0006] The technical scheme of the application is as follows:

[0007] According to a first aspect of the present invention, a method for constructing an integrated energy system optimal scheduling model considering uncertain carbon trading is provided, comprising: predicting a historical carbon price sequence to obtain a predicted carbon price; establishing an uncertain carbon trading model based on the predicted carbon price; establishing an improved conditional value-at-risk model based on uncertain carbon trading based on the uncertain carbon trading model based on the predicted carbon price; and establishing an integrated energy system optimal scheduling model considering uncertain carbon trading based on the uncertain carbon trading model based on the predicted carbon price and the improved conditional value-at-risk model based on uncertain carbon trading.

[0008] Further, the process of predicting the obtained historical carbon price sequence to obtain the predicted carbon price specifically involves: decomposing the historical carbon price sequence using a fully adaptive noise ensemble empirical mode algorithm to obtain the high-frequency components, low-frequency components, and trend term of the historical carbon price; based on the high-frequency components of the historical carbon price, using an exponential generalized autoregressive conditional heteroscedasticity prediction model to obtain the predicted value of the high-frequency components of the carbon price; based on the low-frequency components and trend term of the historical carbon price, using a long short-term memory network prediction model to obtain the predicted value of the low-frequency components and trend term of the carbon price; and superimposing the predicted values ​​of the high-frequency components, low-frequency components, and trend term of the carbon price to obtain the predicted carbon price.

[0009] Furthermore, based on the Pearson correlation coefficient method, the several IMFs obtained by the fully adaptive noise set empirical mode algorithm are divided into high-frequency components and low-frequency components. Specifically, if the Pearson correlation coefficient of an IMF is less than or equal to 0.5, it is classified as a high-frequency component; otherwise, if the Pearson correlation coefficient of an IMF is greater than 0.5, it is classified as a low-frequency component.

[0010] Furthermore, the uncertain carbon trading model based on carbon price prediction is specifically as follows:

[0011]

[0012] In the formula: C CET Carbon trading costs for integrated energy systems; p CET ΔE is the carbon trading price; T is the scheduling period; ΔE t Let a be the amount of carbon allowance used in carbon trading at time t; t The weighted coefficient for predicting carbon prices at time t; Predict the carbon price at time t; Let t be the net carbon emissions of the integrated energy system. Let t be the amount of carbon quota allocated free of charge to the integrated energy system at time t; υ is the carbon quota benchmark coefficient. Let be the output of the i-th gas turbine GT at time t; Let N be the heat output of the i-th gas-fired boiler GB at time t;G is the number of gas turbines GT / gas boilers GB; is the net carbon emission of the integrated energy system at time t; is the total carbon emission of the gas turbine GT and the gas boiler GB at time t; is the carbon emission intensity of the gas equipment; is the volume conversion coefficient; is the molar mass of CO2; is the output of the i-th gas turbine GT at time t; is the heat production power of the i-th gas boiler GB at time t; is the amount of CO2 captured by the carbon capture equipment at time t; is the carbon capture level at time t.

[0013] Further, the improved conditional value at risk model based on uncertain carbon trading is specifically:

[0014]

[0015] In the formula: is the maximum possible risk loss value in the future scheduling period under the confidence level β; c is the carbon trading risk cost; R represents the set; is the probability density function of the predicted carbon price at time t; is the predicted carbon price at time t; is the net carbon emission of the integrated energy system at time t; is the improved conditional value at risk under the confidence level β; is represented by a related loss function.

[0016] Further, the integrated energy system optimization scheduling model considering uncertain carbon trading takes the minimum of the integrated energy system operation cost, the integrated energy system purchase cost of energy from the upper energy, the carbon trading cost of the integrated energy system, the CO2 storage and transportation cost generated by the integrated energy system, and the conditional risk cost of carbon trading as an objective function, and constructs a constraint condition; wherein the constraint condition includes an integrated energy system power balance constraint, an integrated energy system energy storage constraint, a unit capacity and start-stop constraint, a carbon capture equipment constraint, a carbon trading price weighting constraint, and an improved conditional value at risk constraint.

[0017] According to the second aspect of the present application, a method for solving an integrated energy system optimization scheduling model considering uncertain carbon trading is provided, which is used to solve the integrated energy system optimization scheduling model considering uncertain carbon trading constructed according to the method for constructing an integrated energy system optimization scheduling model considering uncertain carbon trading.

[0018] Further, for the improved conditional value at risk model based on uncertainty carbon trading, a virtual variable is introduced to discretize the integral term;A quadratic programming hybrid algorithm based on cross genetic mechanism is used to solve the comprehensive energy system optimization scheduling model considering uncertainty carbon trading, and the operation scheme of the comprehensive energy system optimization scheduling considering uncertainty carbon trading is obtained.

[0019] Further, the quadratic programming hybrid algorithm based on cross genetic mechanism is used to solve the comprehensive energy system optimization scheduling model considering uncertainty carbon trading, and the operation scheme of the comprehensive energy system optimization scheduling considering uncertainty carbon trading is obtained.

[0020] Step 1: taking decision variables As a population individual, an initial population is randomly generated, and the predicted carbon price and known operation data of the comprehensive energy system are introduced;Wherein, a t is the predicted carbon price weighting coefficient at t time; is the carbon capture level at t time;

[0021] Step 2: calculate the fitness of each individual in the population, judge whether the fitness of the individual meets the convergence criterion, and if the convergence criterion is not met, execute step 3;

[0022] Step 3: select parents;The chromosomes of the parents are crossed and genetically evolved to produce offspring, and the offspring chromosomes are mutated to produce new individuals with higher fitness;

[0023] Step 4: the new individuals screened are brought into the comprehensive energy system optimization scheduling model considering uncertainty carbon trading, the quadratic programming problem of the comprehensive energy system optimization scheduling model considering uncertainty carbon trading is converted into a linear programming problem, the solver is called to solve the objective function, and step 2 is returned until the maximum iteration number is reached.

[0024] According to the third aspect of the application, a comprehensive energy system optimization scheduling model considering uncertainty carbon trading is provided, which comprises a processor and a memory, wherein the processor realizes the comprehensive energy system optimization scheduling model considering uncertainty carbon trading when the computer program stored in the memory is executed.

[0025] The beneficial effects of the present application are:

[0026] (1) The CEEMDAN-EGARCH-LSTM combined prediction model is used to predict the carbon price, and the verification shows that the combined prediction model has higher prediction accuracy than other basic models, provides visual information of future short-term carbon market for the comprehensive energy system, and provides transaction basis for uncertainty CET.

[0027] (2) The application fully considers the uncertainty of the carbon transaction participated by the comprehensive energy system, measures the risk level of the carbon transaction based on the conditional risk value of the carbon price, and adds the conditional risk value cost into the objective function of the optimization scheduling model, analyzes the influence of the risk preference coefficient and the confidence level change on the scheduling result. The uncertainty risk of the comprehensive energy system participating in the carbon transaction is quantified by combining the conditional risk value, which can provide certain guidance for the economic and risk of the comprehensive energy system decision maker.

[0028] (3) The predicted carbon price carbon transaction and other types of carbon transactions are compared and analyzed, and the simulation results show that the method has higher superiority in carbon capture level and total operation cost of the comprehensive energy system. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1 The flowchart of the application is shown in the figure;

[0030] Figure 2 The structure diagram of the LSTM neural unit is shown in the figure;

[0031] Figure 3 The energy flow diagram of the comprehensive energy system is shown in the figure;

[0032] Figure 4 The optimization scheduling architecture of the comprehensive energy system based on the carbon price prediction is shown in the figure;

[0033] Figure 5 The CEEMDAN decomposition diagram of the carbon price is shown in the figure;

[0034] Figure 6 The decomposition and recombination diagram of the original carbon price data is shown in the figure;

[0035] Figure 7 The prediction result diagram of each component combination is shown in the figure;

[0036] Figure 8 The comparison diagram of the prediction results of various models is shown in the figure;

[0037] Figure 9 The optimization scheduling result diagram of the electric energy is shown in the figure;

[0038] Figure 10 The optimization scheduling result diagram of the thermal energy is shown in the figure;

[0039] Figure 11 The ICVaR cost diagram under different risk preferences and confidence levels is shown in the figure;

[0040] Figure 12 The influence diagram of the CET carbon price change on the total cost and ICVaR cost is shown in the figure;

[0041] Figure 13 The algorithm iteration process diagram is shown in the figure. DETAILED DESCRIPTION

[0042] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other in any manner without conflict.

[0043] Embodiment 1: As shown in the following, according to a first aspect of the present embodiment, a method for constructing an optimal scheduling model of a comprehensive energy system considering uncertain carbon trading is provided, comprising: predicting a historical carbon price sequence obtained to obtain a predicted carbon price; establishing an uncertain carbon trading model based on carbon price prediction according to the predicted carbon price; establishing an improved conditional value at risk model based on uncertain carbon trading according to the uncertain carbon trading model based on carbon price prediction; and establishing an optimal scheduling model of a comprehensive energy system considering uncertain carbon trading according to the uncertain carbon trading model based on carbon price prediction and the improved conditional value at risk model based on uncertain carbon trading. Figures 1-13 Further, the prediction of the obtained historical carbon price sequence to obtain the predicted carbon price is specifically: the historical carbon price sequence is decomposed by using a completely adaptive noise ensemble empirical mode algorithm to obtain a historical carbon price high-frequency component, a historical carbon price low-frequency component and a historical carbon price trend item; the historical carbon price high-frequency component is predicted by using an exponential generalized autoregressive conditional heteroscedasticity prediction model to obtain a predicted value of the carbon price high-frequency component; the historical carbon price low-frequency component and the historical carbon price trend item are predicted by using a long short-term memory network prediction model to obtain predicted values of the carbon price low-frequency component and the carbon price trend item; and the predicted value of the carbon price high-frequency component, the predicted values of the carbon price low-frequency component and the carbon price trend item are superimposed to obtain the predicted carbon price.

[0044] Further, based on a Pearson correlation coefficient method, a plurality of IMFs obtained by the completely adaptive noise ensemble empirical mode algorithm are divided into high-frequency components and low-frequency components, specifically: if the Pearson correlation coefficient of an IMF is less than or equal to 0.5, the IMF is divided into a high-frequency component; otherwise, if the Pearson correlation coefficient of the IMF is greater than 0.5, the IMF is classified as a low-frequency component.

[0045] Further, when there are multiple components of the historical carbon price high-frequency component, the historical carbon price high-frequency component is reorganized and then predicted. When there are multiple components of the historical carbon price low-frequency component, the historical carbon price low-frequency component is reorganized and then predicted.

[0046] Further, the uncertain carbon trading model based on carbon price prediction is specifically:

[0047]

[0048] ​

[0049]

[0050] C CET is the carbon trading cost of the integrated energy system; p CET is the carbon trading price; T is the dispatch cycle; ΔE t is the carbon quota amount participating in carbon trading at time t; a t is the predicted carbon price weighting coefficient at time t; is the predicted carbon price at time t; is the net carbon emission amount of the integrated energy system at time t; is the amount of carbon quota obtained by the integrated energy system at time t without charge; υ is the carbon quota reference coefficient; is the output of the i-th gas turbine GT at time t; is the heat output of the i-th gas boiler GB at time t; N G is the number of gas turbines GT / gas boilers GB; is the net carbon emission amount of the integrated energy system at time t; is the total carbon emission amount of the gas turbine GT and the gas boiler GB at time t; is the carbon emission intensity of the gas equipment; is the volume conversion coefficient; is the molar mass of CO2; is the output of the i-th gas turbine GT at time t; is the heat output of the i-th gas boiler GB at time t; is the amount of CO2 captured by the carbon capture equipment at time t; is the carbon capture level at time t.

[0051] Further, the improved conditional value at risk model based on uncertainty carbon trading is specifically:

[0052]

[0053] wherein: is the maximum possible risk loss value under the confidence level β in the future dispatch cycle; c is the carbon trading risk cost; R represents the set; is the probability density function of the predicted carbon price at time t; is the predicted carbon price at time t; is the net carbon emission amount of the integrated energy system at time t; is the improved conditional value at risk under the confidence level β; is the loss function.

[0054] ​Further, the comprehensive energy system optimization scheduling model considering uncertain carbon trading takes the minimum of the comprehensive energy system operation cost, the comprehensive energy system purchase cost of energy from the upper energy, the carbon trading cost of the comprehensive energy system, the CO2 storage and transportation cost generated by the comprehensive energy system, and the conditional risk cost of carbon trading as the objective function; and constructs constraint conditions; wherein the constraint conditions include comprehensive energy system power balance constraints, comprehensive energy system energy storage constraints, unit capacity and start-stop constraints, carbon capture device constraints, carbon trading price weighting constraints, and improved conditional value at risk constraints.

[0055] According to a second aspect of the embodiment, a method for solving a comprehensive energy system optimization scheduling model considering uncertain carbon trading is provided, which is used to solve the comprehensive energy system optimization scheduling model considering uncertain carbon trading constructed according to the method for constructing a comprehensive energy system optimization scheduling model considering uncertain carbon trading.

[0056] Further, for the improved conditional value at risk model based on uncertain carbon trading, a virtual variable is introduced to discretize the integral term; a quadratic programming hybrid algorithm based on a cross genetic mechanism is used to solve the comprehensive energy system optimization scheduling model considering uncertain carbon trading, so as to obtain a comprehensive energy system optimization scheduling operation scheme considering uncertain carbon trading.

[0057] Further, the quadratic programming hybrid algorithm based on the cross genetic mechanism is used to solve the comprehensive energy system optimization scheduling model considering uncertain carbon trading, including:

[0058] Step 1: taking decision variables as the population individuals, randomly generating an initial population, and importing the predicted carbon price and known operation data of the comprehensive energy system; wherein, a t is the predicted carbon price weighting coefficient at t moment; is the carbon capture level at t moment;

[0059] Step 2: calculating the fitness of each individual in the population, judging whether the fitness of the individual meets the convergence criterion, and executing Step 3 in the case that the convergence criterion is not met;

[0060] Step 3: selecting parents; performing cross genetic operation on the chromosomes of the parents to generate offspring, and performing mutation on the chromosomes of the offspring to generate new individuals with higher fitness;

[0061] Step 4: using the screened new individuals to bring into the comprehensive energy system optimization scheduling model considering uncertain carbon trading, realizing the conversion of the quadratic programming problem of the comprehensive energy system optimization scheduling model considering uncertain carbon trading into a linear programming problem, calling a solver to solve the objective function, and returning to Step 2 until the maximum number of iterations is reached.

[0062] According to a third aspect of this embodiment, a device for solving an integrated energy system optimization scheduling model considering uncertain carbon trading is provided, comprising a processor and a memory, wherein the processor executes a computer program stored in the memory to implement the method for solving an integrated energy system optimization scheduling model considering uncertain carbon trading as described above.

[0063] According to a fourth aspect of this embodiment, a computer-readable storage medium is provided for storing a computer program, wherein the computer program, when executed by a processor, implements a method for solving an integrated energy system optimization scheduling model considering uncertain carbon trading as described in any one of the above embodiments.

[0064] Example 2: Figures 1-13 As shown, according to a first aspect of this embodiment, a method for constructing an integrated energy system optimal scheduling model considering uncertain carbon trading is provided, including: Step 1, predicting the obtained historical carbon price sequence to obtain the predicted carbon price; Step 2, establishing an uncertain carbon trading model based on the predicted carbon price; Step 3, establishing an improved conditional value-at-risk model based on uncertain carbon trading based on the uncertain carbon trading model based on the predicted carbon price; Step 4, establishing an integrated energy system optimal scheduling model considering uncertain carbon trading based on the uncertain carbon trading model based on the predicted carbon price and the improved conditional value-at-risk model based on uncertain carbon trading. Figure 3 As shown, the integrated energy system contains various energy conversion devices, encompassing the conversion relationships between different energy sources to meet load demands. Electricity sources include gas turbines (GT), grid power, wind power (WT), photovoltaic (PV), and organic Rankine cycle (ORC) waste heat power generation, with electric storage (ES) supplementing the electricity demand. Heat energy primarily comes from ORC waste heat boilers, gas boilers (GB), and electric boilers (EB), with heat storage (HS) supplementing the heat load demand. Gas equipment (GT, GB) is stably supplied by an external gas source, ensuring stable and reliable system operation. Furthermore, although gas equipment reduces carbon emissions to some extent compared to coal-fired equipment, it still falls short of meeting the low-carbon operation requirements of the integrated energy system; therefore, carbon capture equipment is installed for carbon recovery.

[0065] Further, the Step 1 comprises: decomposing the historical carbon price sequence by using a complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) to obtain a high-frequency component, a low-frequency component and a trend term of the historical carbon price; predicting the high-frequency component of the carbon price by using an exponential generalized autoregressive conditional heteroskedasticity (EGARCH) prediction model according to the high-frequency component of the historical carbon price to obtain a predicted value of the high-frequency component of the carbon price; predicting the low-frequency component and the trend term of the carbon price by using a long short-term memory (LSTM) prediction model according to the low-frequency component and the trend term of the historical carbon price to obtain a predicted value of the low-frequency component and the trend term of the carbon price; and superimposing the predicted value of the high-frequency component of the carbon price and the predicted value of the low-frequency component and the trend term of the carbon price to obtain a predicted carbon price.

[0066] Specifically,

[0067] I. Description of CEEMDAN

[0068] The empirical mode decomposition in signal science decomposes a signal into a plurality of different frequency components, i.e., intrinsic mode functions (IMFs), and finally leaves a trend term, which is usually used to decompose complex time-varying and nonlinear signals. Taking the carbon price as the original data, the model thereof can be described as the superposition of a plurality of IMFs and a residual term:

[0069]

[0070] In the formula, x t is a sequence of the original carbon price with respect to time t; F i,t is the i-th IMF of the decomposition, and i = 1,...,n; r t is the final trend term. First, the extreme points of the sample carbon price data are identified, and a cubic spline interpolation method is used to connect all the extreme maximum points to fit an upper envelope line, and the same method is used to fit a lower envelope line. The average of the upper and lower envelope lines is:

[0071]

[0072] In the formula, m t is the average of the upper and lower envelope lines; is the upper envelope line; is the lower envelope line. According to the above formula, the intermediate carbon price data w t can be obtained, and w tFor an IMF component, the expression is:

[0073] w t = x t - m t

[0074] wherein: x t is the original carbon price sequence with respect to time t; m t is the average of the upper and lower envelope lines. The new sequence is the original signal x t and the difference r t of the obtained IMF w t , let r t replace x t and repeat the above steps.

[0075] r i,t = x t - w t

[0076] In order to reduce the decomposition error and the corresponding calculation amount and calculation time, the basic decomposition is improved into the CEEMDAN decomposition algorithm, which is specifically: after each decomposition to obtain an IMF, a Gaussian white noise is added to the residual sequence, the migration of white noise from high frequency band to low frequency band is blocked, and the accurate restoration and complete reservation of the reconstructed signal sequence are ensured. The sample carbon price data x t is added with a Gaussian white noise n i,t to obtain a new data x i,t .

[0077] x i,t = x t + n i,t

[0078] wherein: x i,t is the carbon price sequence with respect to time t after adding the Gaussian white noise for the i-th time; x t is the original carbon price sequence with respect to time t; n i,t is the Gaussian white noise added for the i-th time.

[0079] The x i,t is respectively subjected to empirical mode decomposition to obtain a group of intrinsic mode function components and a trend item, then the average of each decomposition result is taken to obtain the final decomposition result, and according to the vibration frequency range, the high frequency component and the low frequency component are divided:

[0080]

[0081] wherein: F i,t is a historical carbon price component; F i,j,t , r i,n,t are the j-th IMF and the trend item obtained by adding the Gaussian white noise for the i-th time; F th is the high frequency component of historical carbon price; F t l is the low frequency component of historical carbon price; x t is the original carbon price sequence about time t; m is the number of added noise; K is the total number of IMFs decomposed, and k is the number of high frequency components.

[0082] Secondly, based on the Pearson correlation coefficient method, a plurality of IMFs obtained by the CEEMDAN decomposition can be divided into high frequency components and low frequency components. Specifically, the Pearson correlation coefficients of a plurality of IMFs and the original carbon price are calculated respectively, and when the Pearson correlation coefficient of an IMF is less than or equal to 0.5, the IMF is divided into a high frequency component; and when the Pearson correlation coefficient of the IMF is greater than 0.5, the IMF is classified as a low frequency component.

[0083] Thirdly, the description of the exponential generalized autoregressive conditional heteroscedasticity prediction model

[0084] The GARCH model can capture the volatility clustering of the time series, filter high frequency noise, and distinguish between real volatility and random noise. Therefore, the GARCH model is suitable for predicting high frequency modes in the time series and can capture the obvious volatility characteristics in the high frequency components. The GARCH(p, q) model can be modeled as:

[0085]

[0086] In the formula: is the conditional variance; α0, α t is a constant coefficient; γ t is a constant coefficient; p and q are the orders of the GARCH error term and the variance term; ε t-p and σ t-q are the EGARCH error term and the variance term. The general GARCH(1, 1) can describe the volatility clustering effect of a large amount of time series data.

[0087] When the carbon price high frequency sequence exhibits leverage effect and random error does not satisfy the Gaussian distribution, the standard GARCH model can be improved to the EGARCH model. The first-order EGARCH model used in the present application is as follows:

[0088]

[0089] In the formula: is the conditional variance; α0, α1, λ1, β1 are constant coefficients; ε t-1 and σ t-1 are the EGARCH error term and the variance term.

[0090] The improvement of the EGARCH model to the benchmark GARCH model is that: no matter whether ln Take the value, all have Therefore, there is no restriction on all parameters in the model. Now we can give the mean equation of the EGARCH model:

[0091]

[0092] In the formula: is the prediction value of the EGARCH model for the high-frequency component of carbon price; u t is the EGARCH autoregressive part; ε t is the error term; c, φ1, φ2, φ t-i are all undetermined parameters; y t-1 , y t-2 , y t-i is the observation value at the current time point; ξ t is a sampling value conforming to any random distribution. The parameters of EGARCH are determined by the maximum likelihood estimation method.

[0093] Four, LSTM is an improved architecture of recurrent neural network, which introduces unit state to save time series dependence information. In order to effectively control the state, LSTM adopts triple gate mechanism: forget gate, input gate and output gate, and the neural unit structure diagram is shown in Figure 1 . By adding the gating system inside the neuron, the gradient problem that may occur in the calculation of recurrent neural network is solved, and the complex nonlinear mapping between the trend of volatility signal can be learned. Therefore, LSTM can obtain the subtle dependence relationship between data, and further capture and predict the IMF low-frequency and periodic characteristics of carbon price data decomposed by CEEMDAN.

[0094] The forget gate uses the Sigmoid activation function to process the current input and the hidden state at the previous time, generating a gate value in the interval [0, 1] to determine the proportion of information retained by the unit state at the previous time to the current time.

[0095] f t = σ(W f · [h t-1 , x t ]+b f )

[0096] In the formula: f t is the output of the forget gate; σ is the Sigmoid activation function, which compresses the output value to between 0 and 1; W f is the weight matrix of the forget gate; h t-1 is the hidden state at the previous time; x t is the current input low-frequency data; b f is the bias term of the forget gate.

[0097] The input gate filters the effective information of the current input and hidden state using a sigmoid activation function, and the tanh layer updates the candidate memory vector and the unit state.

[0098]

[0099] wherein: i t is the output of the input gate; W i is the weight matrix of the input gate, i = 1, …, n; h t-1 is the hidden state at the previous time step; e t is the current input low-frequency data; b i is the bias term of the input gate, i = 1, …, n; is the new candidate memory value generated by the candidate memory cell; tanh is the hyperbolic tangent activation function, which compresses the output value to between -1 and 1; W C is the weight matrix of the candidate memory cell; b C is the bias term of the candidate memory cell.

[0100] The forgotten information is determined by the forget gate, and a new candidate value is added to form a new unit state belonging to this time step.

[0101]

[0102] wherein: C t is the cell state at the current time step; f t is the output of the forget gate; C t-1 is the cell state at the previous time step; i t is the output of the input gate; is the new candidate memory value generated by the candidate memory cell.

[0103] The output information is determined by the output gate through the sigmoid function.

[0104]

[0105] wherein: F t pre,l is the output of the output gate for the low-frequency component prediction of the carbon price; σ is the Sigmoid activation function, which compresses the output value to between 0 and 1; W o is the weight matrix of the output gate; e ι is the current input low-frequency data; b o is the bias term of the output gate; h ι-1 is the hidden state at the previous time step; h ι is the hidden state at the current time step. o ι is the output coefficient; C ιis the cell state of the current time step. Similarly, the output of the trend item prediction value can be obtained. The carbon price sequence is decomposed into high-frequency, low-frequency and trend items by CEEMDAN, and the EGARCH and LSTM models can be predicted according to the component characteristics. Finally, the predicted components are superimposed to obtain the final predicted carbon price:

[0106]

[0107] In the formula: is the predicted carbon price at time t; F t pre,h is the predicted value of the high-frequency component of the carbon price by the EGARCH model; F t pre ,l is the predicted output of the low-frequency component of the carbon price by the output gate; r t pre is the predicted value of the trend item.

[0108] Further, the Step 2 is specifically: the CO2 generated by the gas equipment (including GT and GB) in the working process is first collected by the carbon capture equipment, and the difference between the carbon production and the carbon capture is the actual carbon emission into the atmosphere, that is, the net carbon emission of the integrated energy system. The carbon emission accounting of the gas equipment and the modeling of the carbon capture equipment are as follows:

[0109]

[0110]

[0111] In the formula: is the total carbon emission of the gas turbine GT and the gas boiler GB at time t; N G is the number of gas turbines GT / gas boilers GB (the number of gas turbines GT and the number of gas boilers GB are equal); is the carbon emission intensity of the gas equipment; is the volume conversion coefficient; is the molar mass of CO2; is the output of the i-th gas turbine GT at time t; is the heat output of the i-th gas boiler GB at time t; is the amount of CO2 captured by the carbon capture equipment at time t; is the carbon capture level at time t; P t cc is the power of the carbon capture equipment at time t; ω c is the electric carbon conversion coefficient; λ cc is the operating efficiency of the carbon capture equipment; is the net carbon emission of the integrated energy system at time t.

[0112]

[0113] C CET is the carbon trading cost of the integrated energy system; p CET is the carbon trading price; T is the dispatch cycle; ΔE t is the carbon quota amount participating in carbon trading at t moment; is the predicted carbon price at t moment; a t is the predicted carbon price weighting coefficient at t moment; is the net carbon emission amount of the integrated energy system at t moment; υ is the carbon quota reference coefficient, the value is 0.7; is the free distributed carbon quota amount of the integrated energy system at t moment; is the output of the i-th gas turbine GT at t moment; is the heat production power of the i-th gas boiler GB at t moment; the weighting coefficient of the predicted carbon price is taken within 0 to 1, the sum of all weighting coefficients is 1, so that the uncertain carbon price value is still within the range of the predicted carbon price, thereby preventing the influence of the predicted extreme value on the carbon trading cost. Finally, the weighting coefficient of the predicted carbon price is used as a decision variable, which only needs to be solved by algorithm iteration to optimize the carbon trading result, reducing carbon emissions while weakening the influence of carbon price uncertainty on the system.

[0114] Further, the Step3 comprises:

[0115] S3.1, establishing a traditional conditional value at risk model CVaR, specifically:

[0116] Let f(x, y) represent a loss function related to the decision vector x, x∈X (X represents the feasible set of the decision variable x), and y∈Y represents a random vector of uncertain factors. Assuming that the probability density function of the random variable is p(y), for a fixed x, the probability of f(x, y) not exceeding a certain loss level a with respect to y is:

[0117] ψ(x, α) = ∫ f(x,y)≤α p(y)dy

[0118] α β (x) = min{α∈R: ψ(x, α) ≥ β}

[0119]

[0120] In the formula, ψ(x, α) is the loss cumulative distribution function under the decision vector x; x is the decision vector; α is a certain loss level; p(y) is the probability density function of the random vector y; y is the random vector; f(x, y) is a loss function related to the decision vector x; α β (x) is the risk value under the confidence level; β is the confidence; Λ β(x) is the conditional value at risk under confidence; R represents a set. Where x is a decision variable, and y is a random variable.

[0121] S3.2, an improved conditional value at risk model based on uncertain carbon trading is established, specifically:

[0122] The improved conditional value at risk model ICVaR based on uncertain carbon trading of the application is jointly influenced by carbon emissions of a comprehensive energy system and predicted carbon price, and a loss function f(x, y) can be changed to That is, the improved conditional value at risk model ICVaR based on uncertain carbon trading is Greater than or equal to The carbon price uncertainty is sampled, and the probability density function of the carbon price is Therefore, the mathematical expression of the improved conditional value at risk model ICVaR based on uncertain carbon trading is given in integral form.

[0123]

[0124] In the formula, is the maximum possible risk loss value in a future scheduling period under a confidence level β, the confidence level is set to 95%; c is a carbon trading risk cost; R represents a set; is the probability density function of the predicted carbon price at time t; is the predicted carbon price at time t; is the net carbon emission of the comprehensive energy system at time t; is the improved conditional value at risk under the confidence level β;

[0125] In the above formula, there is an integral calculation, which cannot be directly solved in a mixed integer linear problem, and a virtual variable z is introduced n The integral term is discretized to obtain The discretization processing is as follows: taking N samples of the predicted carbon price, n = 1, 2, 3,..., N, After simplifying the formula, it is as shown in the following formula:

[0126]

[0127] In the formula, is the improved conditional value at risk under the confidence level β; is the maximum possible risk loss value in a future scheduling period under a confidence level β; p n is the probability density function of the predicted carbon price in the nth sampling; z n is a virtual variable; C CET is the carbon trading cost of the comprehensive energy system; Predicting carbon price at t time; ΔE n Carbon quota amount participating in carbon trading at the nth sampling; The nth sampling predicted carbon price (i.e. the predicted carbon price at the time corresponding to the nth sampling).

[0128] Further, the Step 4 is specifically: considering the comprehensive energy system optimization scheduling model of uncertain carbon trading to minimize the objective function of the comprehensive energy system operation cost, the comprehensive energy system purchasing energy cost from the upper energy, the carbon trading cost of the comprehensive energy system, the CO2 storage and transportation cost generated by the comprehensive energy system, and the conditional risk cost of carbon trading; and constructing the constraint condition; the overall objective function of the comprehensive energy system is:

[0129]

[0130] In the formula: C is the total cost of the comprehensive energy system; Ct is the t time comprehensive energy system operation cost; Ct is the t time comprehensive energy system purchasing energy cost from the upper energy; C CET Ct is the carbon trading cost of the comprehensive energy system; C ST Ct is the CO2 storage and transportation cost generated by the comprehensive energy system; C ICVaR Ct is the conditional risk cost of carbon trading; c GT Ct is the GT operation cost; N G Ct is the number of gas turbines GT / gas boilers GB (the number of gas turbines GT and the number of gas boilers GB); Ct is the t time output of the i th gas turbine GT; Ct is the t time heat power of the i th gas boiler GB; c GT,on-off , c GB,on-off Ct is the start-stop cost of GT and GB respectively; Ct is the start-stop variable of GT and GB at t time, 1 is in operation, and 0 is out of operation; c ORC Ct is the organic Rankine cycle ORC operation cost; P t ORC Ct is the t time organic Rankine cycle ORC output; c EB Ct is the electric boiler EB operation cost; P t EB Ct is the t time electric boiler EB operation power consumption; c WT Ct is the wind power operation cost; P t WT Ct is the t time wind power generation power; c PV Ct is the photovoltaic operation cost; P t PV Ct is the t time photovoltaic power generation power; c ES Ct is the electric energy storage operation cost; P tES is the charging and discharging power of the electric energy storage at time t; c HS is the operating cost of the thermal storage tank; is the charging and discharging power of the thermal storage tank at time t; is the electricity purchase price of the integrated energy system from the upper-level power grid at time t; is the electricity purchase quantity of the integrated energy system from the upper-level power grid at time t; is the electricity sale quantity of the integrated energy system to the upper-level power grid at time t; p gas is the gas purchase price of the integrated energy system from the gas source; is the electricity sale price of the integrated energy system to the upper-level power grid at time t; λ GTe is the power generation efficiency of the GT; λ GBh is the heating efficiency of the GB; G gas is the heating value of natural gas; c st is the storage and transportation cost of capturing one unit of CO2; is the quantity of CO2 captured by the carbon capture device at time t; is the improved conditional value at risk at a confidence level β; r is a risk preference coefficient, and the more conservative the risk preference coefficient is, the larger the risk preference coefficient is determined by the decision maker, and the present application is set to 0.5.

[0131] The constraint conditions include: integrated energy system power balance constraint, integrated energy system energy storage constraint, unit capacity and start-stop constraint, carbon capture device constraint, carbon trading price weighted constraint, and improved conditional value at risk constraint.

[0132] The integrated energy system power balance constraint is:

[0133]

[0134] The integrated energy system energy storage constraint is:

[0135]

[0136] The unit capacity and start-stop constraint is:

[0137]

[0138] The carbon capture device constraint is:

[0139]

[0140] The carbon trading price weighted constraint is:

[0141]

[0142] The improved conditional value at risk constraint is:

[0143]

[0144] In the formula: The amount of electricity purchased by the integrated energy system from the superior power grid at time t; P represents the amount of electricity sold by the integrated energy system to the upper-level power grid at time t. t WT P represents the wind power generation capacity at time t. t PV Let t be the photovoltaic power generation at time t; Let P be the power output of the i-th gas turbine GT at time t; t cc P represents the power of the carbon capture device at time t. t ORC,e P represents the waste heat power generation capacity of the Organic Rankine Cycle (ORC) at time t. t EB Let t be the power consumed by the electric boiler EB during operation; The heat output of the Organic Rankine Cycle (ORC) waste heat boiler; Let GB be the heat output of the i-th gas-fired boiler at time t; P represents the heat output of the electric boiler. t ES,ch H t HS,ch These represent the charging power of the electrical energy storage ES and the thermal storage tank HS at time t, respectively; P t ES,dis H t HS,dis Let HS and the power released by the thermal storage tank be at time t, respectively. These represent the electricity and heat load demands at time t, respectively; P t ES Let λ be the energy storage charge / discharge power at time t; ES,loss , λ HS,loss λ represents the self-loss coefficients of the electrical energy storage (ES) and the thermal energy storage tank, respectively. ES,ch , λ HS,ch λ represents the charging efficiency of the electric energy storage (ES) and the thermal storage tank, respectively. ES,dis , λ HS,dis The discharge efficiencies of ES and the thermal storage tank are respectively; P ES,max H HS,max These represent the capacities of the electrical energy storage (ES) and the thermal energy storage tank (HS); ΔP ES,ch ΔH HS,ch These represent the maximum charging power of the electrical energy storage ES and the thermal storage tank HS at time t, respectively; ΔP ES,dis ΔH HS,dis These represent the maximum output power of the electric energy storage ES and the thermal storage tank at time t, respectively. P represents the state of charge of ES and the thermal storage tank, respectively, and is a 0-1 variable; t WT,pre P represents the predicted wind power at time t.t PV,pre is the photovoltaic predicted power at time t; λ EB is the electric boiler EB heating efficiency; P EB,max is the EB maximum heating power; is the i-th gas turbine GT heat generation power at time t; P t ORC is the organic Rankine cycle ORC output at time t; is the i-th gas turbine GT output at time t; P ORC,max is the ORC waste heat power generation maximum power; are the waste heat power generation ramp-up and ramp-down limits, respectively; a is the waste heat recovery distribution coefficient; λ GTh is the GT heating efficiency; r WHB is the waste heat boiler recovery efficiency; r ORC is the waste heat power generation efficiency; b t GT , b t GB are the GT and GB start-stop variables at time t, 1 for operation and 0 for shutdown; λ GB is the GB heat production efficiency; P GT,max , P GT,min are the GT maximum and minimum rated powers; H GB,max , H GB,min are the GB maximum and minimum powers; R GT , R GB are the GT and GB ramp rates; and are the GT start-up and shutdown time counters at time t-1; T GT,on and T GT,off are the GT minimum start-up and shutdown durations; and are the GB start-up and shutdown time counters at time t-1; T GB,on and T GB,off are the GB minimum start-up and shutdown durations; P grid,max is the integrated energy system and upper-level power grid tie line constraint; is the carbon capture level at time t; is the maximum carbon capture level of the carbon capture device, taking 1; a t is the predicted carbon price weighting coefficient at time t; z n is a virtual variable, taking values n = 1, 2, 3, …, N; C CET is the carbon trading cost of the integrated energy system; is the maximum possible risk loss value in the future scheduling period under the confidence level β.

[0145] The application can obtain high frequency, low frequency and trend items by decomposing the carbon price sequence through CEEMDAN, and the EGARCH and LSTM models can be used for prediction according to the component characteristics, and the comprehensive energy system can reduce the net carbon emission under the strong constraint of the carbon price; on the premise of accurate prediction of the carbon price, the transaction carbon price beneficial to the comprehensive energy system needs to be determined within the predicted carbon price range, and the application adopts the carbon transaction based on the predicted carbon price weighted value to guide the comprehensive energy system to the scheduling planning of a lower carbon; then, the influence of the carbon transaction uncertainty on the economy of the comprehensive energy system is quantified by improving the conditional value at risk, and the measurement of the CET risk level is realized; finally, the operation cost of the comprehensive energy system, the purchasing cost of the comprehensive energy system to the upper energy, the carbon transaction cost of the comprehensive energy system, the storage and transportation cost of the CO2 generated by the comprehensive energy system, and the conditional risk cost of the carbon transaction are taken as the objective function; and the constraint conditions including the power balance constraint of the comprehensive energy system, the energy storage constraint of the comprehensive energy system, the unit capacity and start-stop constraint, the carbon capture device constraint, the carbon transaction price weighted constraint and the improved conditional value at risk constraint are constructed.

[0146] According to the second aspect of the embodiment, a method for solving the comprehensive energy system optimal scheduling model considering the uncertain carbon transaction is provided, specifically: a YALMIP toolbox is called by MATLAB2024a to construct the comprehensive energy system optimal scheduling model considering the uncertain carbon transaction, a quadratic programming hybrid algorithm based on a cross genetic mechanism is used to solve the comprehensive energy system optimal scheduling model considering the uncertain carbon transaction, and a comprehensive energy system optimal scheduling operation scheme considering the uncertain carbon transaction is obtained.

[0147] It should be noted that in the comprehensive energy system optimal scheduling model, the uncertain carbon price weighted coefficient and the carbon quota transaction amount are decision variables, the model solving is converted into a quadratic programming problem, and a genetic algorithm is usually used to solve this problem. However, in actual application, the genetic algorithm is difficult to deeply mine the internal correlation of the constraint conditions in the model, so that the search process lacks directionality, and the random variation mechanism easily causes the population to converge to a local extreme point too early. Therefore, the application provides a quadratic programming hybrid algorithm based on a cross genetic mechanism, which retains the flexibility of the genetic algorithm in processing complex constraints, and avoids the local optimal trap through the deterministic optimization characteristics of the quadratic programming. The solving process is as follows:

[0148] Step 1: randomly generate an initial population, and import the predicted carbon price and known operation data of the comprehensive energy system; the known operation data includes wind power prediction power, photovoltaic prediction power and electric and thermal load demand. Exemplarily, the decision variables are The initial values of the individuals are obtained in a random manner.

[0149] Step 2: Calculate the fitness of each individual in the population, and determine whether the fitness of the individual meets the convergence criterion |C k -C k-1 |≤ε: if yes, end; otherwise, execute step 3. The fitness function is:

[0150]

[0151] wherein: is the fitness function of the kth iteration; C k is the total cost of the integrated energy system of the kth iteration; is the fitness penalty coefficient of the kth iteration; is the constraint violation of the kth iteration, i.e., the absolute value sum of the individual exceeding the constraints of the optimization scheduling model; and ε is the iteration error, which is set to 0.1.

[0152] In a specific application, C1 is initialized to infinity, and k is set to 1, and C0 is set to 0.

[0153] Step 3: Apply the survival of the fittest principle to select parents, which is specifically: individuals with high fitness are more likely to be selected, and individuals with low fitness are gradually eliminated.

[0154] Step 4: Cross the chromosomes of the parents to produce offspring, and mutate the chromosomes of the offspring to produce new individuals with higher fitness, and k = k + 1.

[0155] Step 5: Use the new individuals selected to bring into the integrated energy system optimization scheduling model considering uncertain carbon trading, to convert the quadratic programming problem of the integrated energy system optimization scheduling model considering uncertain carbon trading into a linear programming problem, and call the Gurobi solver to solve the objective function C k , and strictly follow the operating constraints of the integrated energy system, and return to step 2 until the maximum number of iterations is reached, and output the Pareto optimal solution.

[0156] As can be seen from the above technical solution, this algorithm can use the global search capability of genetic algorithm to avoid falling into local optimal solution, and through the local optimization capability of quadratic programming to perform fine search near the potential feasible solution found by genetic algorithm, thereby quickly and accurately positioning to a high-quality solution, and improving the model optimization capability. The present application constructs a model by calling the YALMIP toolbox in MATLAB, and solves it by using the Gurobi solver.

[0157] Further, the present application selects the carbon trading data of the national carbon emission trading platform from December 9, 2024 to December 14, 2024 as the research object (i.e., the original carbon price x t), the spring operation data of a certain comprehensive energy town in Lanzhou New Area of Gansu Province is selected as the data source to participate in the constraint. The purchase and sale electricity price of the comprehensive energy system adopts time-of-use electricity price, as shown in Table 1.

[0158] Table 1 Purchase and sale electricity price of comprehensive energy system

[0159]

[0160] The CEEMDAN algorithm is used to decompose and process the historical carbon price sequence. In the decomposition calculation process, the iteration number of the algorithm is set to 100 times, the white noise is set to 10% of the standard deviation of the original data, that is, 0.1 times the standard deviation level. Five IMF components and one trend item are obtained in the decomposition process, and the decomposition result is as shown in Figure 5 From the frequency domain characteristics, the high-frequency IMF component has the characteristics of dense extreme point distribution and symmetric data fluctuation. The low-frequency IMF component has the characteristics of sparse extreme point distribution and increasing distribution interval, and the mean value deviates from zero. Based on the difference analysis of IMF frequency domain characteristics, high-frequency group and low-frequency group can be divided. The trend item is the residual output of decomposition, and has the characteristics of global monotonicity or weak volatility, contains the development trend of the original sequence, so it is defined as the trend item to represent the long-term dynamic characteristics of carbon price evolution. Based on the Pearson correlation coefficient of IMF and the original carbon price is 0.5, less than or equal to 0.5 is divided into high-frequency group, and greater than 0.5 is divided into low-frequency group, so IMF1-IMF5 can be grouped, and the standard deviation and Pearson correlation coefficient of three groups including the trend item are calculated, as shown in Table 2.

[0161] Table 2 CEEMDAN component reorganization table

[0162]

[0163] Since the high-frequency group has multiple components, reorganization is carried out, and according to the IMF component reorganization analysis result and the time sequence characteristics of each group data Figure 6 , it can be seen that Figure 6It can be seen that the low frequency group is highly correlated with the original carbon price sequence and highly consistent with the overall trend of the original sequence. The high frequency group shows typical carbon market structure characteristics, which confirm that the high frequency can capture the price fluctuations caused by market order flow shocks. The trend component generally shows a decreasing function, which represents the overall trend of the carbon price market. Based on the EGARCH(1, 1) model, a single-step rolling prediction is performed on the high frequency group, and the carbon price time series prediction value is added as input to generate the prediction sequence of the next 24 time points. For the low frequency group and the trend component, an LSTM neural network prediction model is constructed, with the parameter configuration set to 128 hidden units, 600 training iterations, a gradient threshold of 1, an initial learning rate of 0.08, and a dynamic decay strategy. The comparison of the prediction results and the true values of the test set is shown in Figure 7 The IMF and trend component of different frequency groups have followability to the actual value, and the change trend is more consistent with the dynamic development of the actual carbon price.

[0164] The prediction results of the present application are shown in Figure 7 To evaluate the prediction effect, the prediction evaluation indexes of the present application and other prediction methods are compared, and the results are shown in Table 3.

[0165] Table 3: Accuracy evaluation of each model

[0166] Model MAE / (Yuan / Ton) RMSE / (Yuan / Ton) MAPE / % ARIMA 1.0925 1.3137 1.0995 EGARCH 1.6455 1.9324 0.0661 LSTM 1.1522 1.2560 0.0162 CEEMDAN-EGARCH 1.5184 1.8739 0.0152 CEEMDAN-LSTM 0.9243 1.1075 0.0093 CEEMDAN-EGARCH-LSTM 0.9011 1.0021 0.0051

[0167] From Table 3, it can be seen that the CEEMDAN-EGARCH-LSTM combination prediction model is better than ARIMA. Although the ARIMA model is relatively traditional and widely used, the national carbon trading price is disturbed by its own short-term sequence and long-term fluctuation, and ARIMA cannot capture the fluctuation trend of the carbon price time sequence, so that the prediction error is increased. The CEEMDAN-EGARCH and CEEMDAN-LSTM models are better than the prediction effect of the single EGARCH model and the LSTM model, which shows that after the CEEMDAN model decomposes the carbon price into components of different frequencies, EGARCH and LSTM can capture the fluctuation information of the carbon price, but the prediction accuracy still lags behind the CEEMDAN-EGARCH-LSTM combination prediction model. The CEEMDAN-EGARCH-LSTM combination prediction model proposed in the application can select the corresponding EGARCH model and LSTM model for prediction according to the different frequencies, which can maximize the prediction advantage of the prediction model. The MAE, RMSE and MAPE three precision evaluation indexes are better than other prediction models. Based on the CEEMDAN-EGARCH-LSTM combination prediction model, the differences of different frequencies can be finely decomposed, the complex characteristics of carbon market fluctuation can be effectively captured, and the precision of carbon price prediction is improved, which provides a feasible value basis for the uncertainty of carbon trading.

[0168] To verify the effectiveness and superiority of the comprehensive energy system optimization scheduling model considering the uncertainty of CET proposed in the application, three scenarios are set to compare and analyze the method of the application:

[0169] Scenario 1: Carbon trading CET does not participate in the optimization scheduling of the comprehensive energy system, but carbon tax is levied on carbon emissions in addition to carbon quota;

[0170] Scenario 2: The ladder type CET participates in the optimization scheduling of the comprehensive energy system and considers the ICVaR risk assessment;

[0171] Scenario 3: Uncertain CET participates in the optimization scheduling of the comprehensive energy system and considers the ICVaR risk assessment, the research method of the application.

[0172] The optimization scheduling cost and carbon emission of the three scenarios are compared, as shown in Table 4.

[0173] Table 4 Comparison of optimization scheduling cost and carbon emission of three scenarios

[0174]

[0175] In Table 4, E os That is, the cumulative sum of the scheduling period, and the others are the same.

[0176] ​Scenario 1 is the optimal scheduling of integrated energy system without considering the participation of carbon trading and imposing carbon tax on net carbon emissions, and scenario 3 is the optimal scheduling model of integrated energy system considering the conditional value at risk and the participation of uncertain CET. As can be seen from Table 4, the carbon capture equipment participates in both scenarios 1 and 3. Since scenario 1 does not consider the participation of CET in the optimal scheduling of integrated energy system, the output of carbon capture equipment is reduced, and its carbon capture level is 34%, resulting in a daily net carbon emission of 4510.0 tons in the optimal scheduling of integrated energy system, which is significantly higher than that of scenario 3 considering the uncertain carbon trading of integrated energy system. However, although scenario 1 does not consider the participation of CET, it considers the impact of carbon emissions on the environment, so a carbon tax of 143 yuan / ton is imposed in this scenario, and therefore the total scheduling cost of integrated energy system in scenario 1 is higher.

[0177] The rated capacity of the equipment in scenario 2 and scenario 3 is the same, but due to the different types of carbon trading, the output of the device in the system is different. Compared with scenario 3, scenario 2 only sets the carbon price relative to the system carbon emissions, which can only adjust the output of the gas equipment and the carbon capture equipment slowly, so that the output of the carbon capture equipment in scenario 3 is small, and the carbon capture level is 33.1%. In scenario 3, after predicting the carbon price, the carbon price change is weighted for carbon trading, which can greatly guide the reduction of carbon emissions, and under the optimization and regulation of other types of power supply, the power of the carbon capture equipment is improved, and the carbon capture level of scenario 3 is improved. Therefore, the system carbon emissions are significantly reduced, and the carbon quota obtained is surplus, and the system obtains income by selling the carbon quota. In addition, scenario 3 uses the CEEMDAN-EGARCH-LSTM combined model to accurately predict the carbon price, the carbon price value follows the real carbon price, the risk value cost is reduced, and the total cost is also reduced.

[0178] The results of the optimal scheduling of electric energy are shown in Figure 9 When the sunlight is sufficient, the photovoltaic power is high during the period of 08:00-16:00, which directly meets the demand of electric load and reduces the dependence on GT and external power purchase. During the night or high wind speed period of 20:00-04:00, the wind power contributes significantly, which is complementary to photovoltaic. When the output of photovoltaic and wind power is excessive, the electric energy storage charges, and stores the excess electric energy. However, the electric energy storage still discharges at the time of heavy load. When the self-generation of integrated energy system cannot meet the demand of electric load, the power is purchased from the external main grid, and the cost is affected by the time-of-use electricity price. When new energy is sufficient, wind power and electric energy storage discharge are used preferentially to avoid the full-capacity start of GT with high carbon emissions. When new energy is insufficient, GT is gradually started to balance the economy and low carbon of integrated energy system.

[0179] The results of the optimal scheduling of thermal energy are shown in Figure 10As shown, from 00:00 to 04:00, the heat load is relatively low, mainly met by the heat release from GB and HS. This indicates that during the off-peak hours at night, the system can utilize the heat energy stored in the thermal storage tank to reduce the cost of real-time heat generation. HS also supplements the carbon emission penalty start-up and shutdown of GB. Then, from 04:00 to 08:00, the heat load gradually increases, and EB and ORC begin to increase their output to prepare for the surge in heat load during the morning peak, while adjusting according to energy availability. During the peak period from 08:00 to 12:00, all heat sources work together, especially HS, which releases stored energy, while EB and ORC supplement. From 12:00 to 16:00, the heat load decreases but remains high, and the heat sources continue to coordinate. From 16:00 to 20:00, it rises slowly again, GB outputs heat energy normally, and HS increases its heat storage to store energy for the night. From 20:00 to 24:00, the load decreases, and the output heat energy of each heating device gradually decreases.

[0180] Carbon price fluctuations introduce uncertainty into the participation of integrated energy systems in carbon market trading, leading to risk costs for these systems, such as... Figure 11 As shown, the risk cost varies under different risk appetites and confidence levels. As the risk appetite of integrated energy systems participating in carbon trading becomes more conservative, the optimized ICVaR value decreases slightly. That is, at a certain confidence level, the conditional risk cost C for participating in CET carbon trading based on predicted carbon prices decreases. ICVaR Slightly lower. Meanwhile, when adopting a conservative risk-preference decision-making approach, the carbon emission constraints on the system are reduced due to stricter risk constraints. The integrated energy system will significantly utilize wind and solar turbines or units with lower carbon emissions as output, reducing the operating power of gas-fired equipment and maximizing the operating power of carbon capture equipment. This aligns with the strategy of this invention, which uses an uncertain carbon trading model to optimize the scheduling of the integrated energy system towards a lower carbon footprint, thereby reducing the amount of carbon allowances traded in the carbon market to avoid trading losses. With a fixed risk preference coefficient r, ICVaR shows a trend of first decreasing and then increasing as the confidence level increases, indicating that there is a confidence level corresponding to the minimum risk level. Compared to changes in the risk preference coefficient, changes in confidence level have a more significant impact on the ICVaR value, indicating that confidence level significantly affects the risk level of carbon trading in the system.

[0181] like Figure 12As shown, with the increase of the uncertain carbon price, the total cost of the integrated energy system first increases and then decreases under the solution of the objective function, while the ICVaR decreases. It can be seen that the increase of the carbon price directly increases the CET cost, resulting in the increase of the total cost of the integrated energy system. In this stage, the decrease of the ICVaR value is difficult to offset the increase of the CET cost when the system responds to the carbon price fluctuation, resulting in the passive increase of the total cost. With the carbon price reaching a certain value, the integrated energy system optimizes the equipment operation mode, greatly reduces the carbon emission, and the steepness of the ICVaR value decreases, so that the total cost of the integrated energy system decreases under the solution of the objective function. Further analysis shows that when the carbon trading price is higher than 75 yuan / ton, the integrated energy system shows positive improvement in terms of cost and risk. Higher carbon trading price promotes the system to take more active low-carbon measures, and also reduces the cost and risk caused by carbon emission. In addition, increasing the risk preference coefficient can also reduce the transaction risk and the total operation cost of the integrated energy system.

[0182] The cross genetic quadratic programming algorithm and the traditional genetic algorithm are used to solve the scheduling model of the integrated energy system, and the total cost iterative process is compared as shown in the figure. Figure 13 As shown, the cross genetic quadratic programming algorithm reaches the iterative convergence value in a smaller number of iterations, and the final convergence value is 1667921.97 yuan, which is 100.13 yuan lower than that of the traditional genetic algorithm, and the solution of the cross genetic quadratic programming algorithm is more stable and concentrated. Compared with the traditional genetic algorithm, the cross genetic quadratic programming algorithm has optimized convergence speed, optimization time and algorithm convergence.

[0183] It can be known from the above technical solutions that the method improves the carbon price prediction accuracy, and improves the economy and carbon emission level of the integrated energy system.

[0184] The specific embodiments of the application are described in detail above in combination with the drawings, but the application is not limited to the above embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the purpose of the application.

Claims

1. A method for constructing an integrated energy system optimal scheduling model considering uncertain carbon trading, characterized in that, include: The obtained historical carbon price series is used to predict the predicted carbon price. Based on predicted carbon prices, establish an uncertainty carbon trading model based on carbon price predictions; Based on the uncertain carbon trading model based on carbon price forecasting, an improved conditional value-at-risk model for uncertain carbon trading is established. Based on the uncertain carbon trading model based on carbon price forecasting and the improved conditional value-at-risk model based on uncertain carbon trading, a comprehensive energy system optimization scheduling model considering uncertain carbon trading is established.

2. The method for constructing an integrated energy system optimal scheduling model considering uncertain carbon trading as described in claim 1, characterized in that, The process of predicting the obtained historical carbon price series to obtain the predicted carbon price specifically involves: The historical carbon price sequence is decomposed using a fully adaptive noise ensemble empirical mode algorithm to obtain the high-frequency components, low-frequency components, and trend term of the historical carbon price. Based on the high-frequency components of historical carbon prices, an exponential generalized autoregressive conditional heteroscedasticity prediction model is used to predict the high-frequency components of carbon prices. Based on the historical low-frequency components and trend terms of carbon prices, a long short-term memory network prediction model is used to make predictions and obtain the predicted values ​​of the low-frequency components and trend terms of carbon prices. The predicted carbon price is obtained by superimposing the predicted values ​​of the high-frequency component of carbon price, the low-frequency component of carbon price, and the trend term.

3. The method for constructing an integrated energy system optimization scheduling model considering uncertain carbon trading as described in claim 2, characterized in that, Based on the Pearson correlation coefficient method, several IMFs obtained by the fully adaptive noise set empirical mode algorithm are divided into high-frequency components and low-frequency components. Specifically, if the Pearson correlation coefficient of an IMF is less than or equal to 0.5, it is classified as a high-frequency component; otherwise, if the Pearson correlation coefficient of an IMF is greater than 0.5, it is classified as a low-frequency component.

4. The method for constructing an integrated energy system optimal scheduling model considering uncertain carbon trading as described in claim 1, characterized in that, The uncertainty-based carbon trading model based on carbon price prediction is specifically as follows: In the formula: C CET Carbon trading costs for integrated energy systems; p CET ΔE is the carbon trading price; T is the scheduling period; ΔE t Let a be the amount of carbon allowance used in carbon trading at time t; t The weighted coefficient for predicting carbon price at time t; Predict the carbon price at time t; Let t be the net carbon emissions of the integrated energy system. Let t be the amount of carbon quota allocated free of charge to the integrated energy system at time t; υ is the carbon quota benchmark coefficient. Let be the output of the i-th gas turbine GT at time t; Let N be the heat output of the i-th gas-fired boiler GB at time t; G For the number of gas turbines (GT) / gas boilers (GB); Let t be the net carbon emissions of the integrated energy system. Let t be the total carbon emissions of the gas turbine GT and the gas boiler GB at time t; Carbon emission intensity of gas-fired equipment; This is the volume conversion factor; This represents the molar mass of CO2. Let be the output of the i-th gas turbine GT at time t; Let GB be the heat output of the i-th gas-fired boiler at time t; Let t be the amount of CO2 captured by the carbon capture device at time t; Let t be the carbon capture level at time t.

5. The method for constructing an integrated energy system optimal scheduling model considering uncertain carbon trading as described in claim 1, characterized in that, The improved conditional value-at-risk model based on uncertain carbon trading is specifically as follows: In the formula: Let be the maximum possible risk loss value within a future scheduling cycle under a confidence level β; c is the carbon trading risk cost; R represents the set. Let be the probability density function for predicting the carbon price at time t; Predict the carbon price at time t; Let t be the net carbon emissions of the integrated energy system. To improve the conditional value of risk at confidence level β; To indicate The relevant loss function.

6. The method for constructing an integrated energy system optimal scheduling model considering uncertain carbon trading as described in claim 1, characterized in that, The integrated energy system optimization scheduling model considering uncertain carbon trading takes the following as its objective function: the integrated energy system's operating cost, the integrated energy system's energy purchase cost from upstream energy sources, the integrated energy system's carbon trading cost, the integrated energy system's CO2 storage and transportation costs, and the conditional risk cost of carbon trading. Constraints are also constructed, including integrated energy system power balance constraints, integrated energy system energy storage constraints, unit capacity and start-up / shutdown constraints, carbon capture equipment constraints, carbon trading price weighted constraints, and improved conditional risk value constraints.

7. A method for solving an integrated energy system optimal scheduling model considering uncertain carbon trading, characterized in that, The integrated energy system optimization scheduling model considering uncertain carbon trading, constructed according to the method for constructing an integrated energy system optimization scheduling model considering uncertain carbon trading as described in claim 1, is solved.

8. The method for solving the integrated energy system optimal scheduling model considering uncertain carbon trading as described in claim 7, characterized in that, For the improved conditional value-at-risk model based on uncertain carbon trading, dummy variables are introduced to discretize the integral term; A quadratic programming hybrid algorithm based on crossover genetics is used to solve the integrated energy system optimization scheduling model considering uncertain carbon trading, and the optimized scheduling operation scheme of the integrated energy system considering uncertain carbon trading is obtained.

9. The method for solving the integrated energy system optimal scheduling model considering uncertain carbon trading as described in claim 8, characterized in that, The method employs a quadratic programming hybrid algorithm based on a crossover genetic mechanism to solve the integrated energy system optimization scheduling model considering uncertain carbon trading, including: Step 1: Using decision variables As individuals within a population, an initial population is randomly generated, and known operational data on predicted carbon prices and integrated energy systems are imported; where a t The weighted coefficient for predicting carbon price at time t; Let t be the carbon capture level; Step 2: Calculate the fitness of each individual in the population and determine whether the fitness of an individual meets the convergence criterion. If the convergence criterion is not met, proceed to Step 3. Step 3: Select parents; perform cross-linking inheritance of the parents' chromosomes to produce offspring, and mutate the offspring's chromosomes to produce new individuals with higher fitness; Step 4: Use the newly selected individuals to enter the integrated energy system optimization scheduling model that considers uncertain carbon trading, so as to transform the quadratic programming problem of the integrated energy system optimization scheduling model that considers uncertain carbon trading into a linear programming problem. Call the solver to solve the objective function, return to step 2, and continue until the maximum number of iterations is reached.

10. A device for solving an integrated energy system optimization scheduling model considering uncertain carbon trading, characterized in that, It includes a processor and a memory, wherein when the processor executes a computer program stored in the memory, it implements a solution method for an integrated energy system optimization scheduling model considering uncertain carbon trading as described in any one of claims 7-9.

Images