Wind-solar-storage station control methods, equipment, media and products considering electricity-green-carbon price forecasts

Abstract

The present application discloses a method, device, medium, and product for regulating a wind-solar-storage station that considers the prediction of electricity-green-carbon prices, and relates to the field of new energy technology. The method includes inputting a combination of acquired historical price data into a price prediction model to output a predicted price; determining a deviation vector of the price data based on the combination of historical price data and the predicted price, and using a multi-core-based single-class support vector machine algorithm to generate an uncertainty set for the predicted price; using a neural network classifier to classify the uncertainty set of the predicted price to obtain multiple price scenarios; using a spotted kingfisher optimization algorithm to solve a joint clearing model based on the predicted prices under multiple price scenarios to obtain an operating strategy for the wind-solar-storage station; and regulating the wind-solar-storage station based on the operating strategy of the wind-solar-storage station. The present application improves the prediction accuracy of electricity prices, carbon sink prices, and green certificate prices, and enhances the robustness of the regulation of wind-solar-storage stations.

Description

Technical Field

[0001] The present application relates to the field of new energy power generation technology, and in particular to a wind-solar-storage station control method, equipment, medium and product that considers electricity-green-carbon price forecasts. Background Art

[0002] With the acceleration of the global energy transition, the proportion of renewable energy electricity generation is increasing year by year. To cope with the uncertainty of renewable energy generation, grid dispatching needs to be more flexible. Furthermore, various price factors, such as electricity prices, carbon sink prices, and green certificate prices, influence the arrangement and implementation of renewable energy station dispatch plans. Currently, electricity prices, carbon sink prices, and green certificate prices all exhibit significant time series characteristics, with complex long-term and short-term dependencies, and are subject to significant volatility and uncertainty. However, traditional forecasting methods struggle to fully capture these characteristics, thus affecting forecast accuracy. Furthermore, given price fluctuations and uncertainty, how to quickly and efficiently obtain the fluctuation range of price forecasts and improve the robustness of wind, solar, and storage station dispatch plans within this range is a key issue that urgently needs to be addressed. Summary of the Invention

[0003] The purpose of this application is to provide a wind, solar and storage station control method, equipment, medium and product that takes into account the electricity-green-carbon price forecast, improve the prediction accuracy of electricity prices, carbon sink prices and green certificate prices, and enhance the robustness of wind, solar and storage station control.

[0004] To achieve the above objectives, this application provides the following solutions.

[0005] In the first aspect, the present application provides a method for obtaining a combination of historical price data; the combination of historical price data includes: historical carbon prices, historical green certificate prices, and historical electricity prices;

[0006] The historical price data combination is input into a price prediction model to output a predicted price; the predicted price includes: a predicted carbon price value, a predicted green certificate price value, and a predicted electricity price value; the price prediction model is obtained by training an LSTM neural network model based on a hybrid attention mechanism using a training data set; the training data set includes a sample price data combination and the corresponding predicted value of the sample price data combination;

[0007] determining a deviation vector of price data based on the combination of the historical price data and the predicted price;

[0008] A multi-core based single-class support vector machine algorithm is used to generate an uncertainty set for the predicted price based on the deviation vector of the price data;

[0009] A neural network classifier is used to classify the uncertainty set of the predicted price to obtain multiple price scenarios; each price scenario includes multiple predicted prices;

[0010] Using the Pied Kingfisher optimization algorithm, the joint clearing model is solved based on the predicted prices under multiple price scenarios to obtain the operation strategy of the wind, solar and storage station. The joint clearing model includes a wind, solar and storage station power generation strategy model, a thermal power unit power generation strategy model, a first objective function, and a first constraint condition. The wind, solar and storage station power generation strategy model includes a second objective function and a second constraint condition. The thermal power unit power generation strategy model includes a third objective function and a third constraint condition.

[0011] The wind-solar-storage station is regulated based on its operation strategy.

[0012] In the second aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-mentioned methods for controlling wind, solar, and storage stations taking into account electricity-green-carbon price forecasts.

[0013] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned methods for controlling wind, solar and storage stations taking into account electricity-green-carbon price forecasts.

[0014] In a fourth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned methods for controlling wind, solar and storage stations taking into account electricity-green-carbon price forecasts.

[0015] According to the specific embodiments provided in this application, the present application has the following technical effects: This application discloses a wind, solar, and storage station control method, equipment, medium, and product that considers electricity, green energy, and carbon price forecasts. A price forecasting model is established to predict carbon prices, green energy certificate prices, and electricity prices. The price forecasting model is based on an LSTM neural network model based on a hybrid attention mechanism, which can effectively capture the long-term and short-term dependencies between electricity prices, carbon sink prices, and green energy certificate prices in time series data, thereby improving the accuracy of price forecasts. A multi-core single-class support vector machine algorithm is used to generate the uncertainty set of the predicted price, effectively processing outliers and noisy data and improving the reliability of the uncertainty set. A neural network classifier is used to classify the uncertainty set of the predicted price, improving the accuracy and stability of price scenario classification. Finally, a joint clearing model for multiple price scenarios is established, including a wind, solar, and storage station generation strategy model and a thermal power unit generation strategy model. This model can achieve efficiency stability under price uncertainty and effectively regulate the operation of the wind, solar, and storage station. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 A flow chart of a wind, solar and storage station control method considering electricity, green energy and carbon price forecasts provided in one embodiment of the present application.

[0018] Figure 2 Another flowchart of a wind, solar and storage station control method considering electricity, green and carbon price forecasts provided in an embodiment of the present application.

[0019] Figure 3 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0021] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0022] In an exemplary embodiment, Figure 1 and Figure 2 As shown, a wind-solar-storage station control method considering electricity-green-carbon price forecast is provided, including the following steps S1 to S7.

[0023] Step S1: Acquire a combination of historical price data; the combination of historical price data includes: historical carbon prices, historical green certificate prices, and historical electricity prices. The carbon price is the price in the carbon trading market, the green certificate price is the price in the green certificate trading market, and the electricity price is the day-ahead electricity price in the electricity spot market.

[0024] Step S2: Input the historical price data combination into a price prediction model and output a predicted price; the predicted price includes: a predicted carbon price, a predicted green certificate price, and a predicted electricity price; the price prediction model is obtained by training an LSTM neural network model based on a hybrid attention mechanism using a training dataset; the training dataset includes a sample price data combination and the corresponding predicted value of the sample price data combination. The sample price data combination is a price data combination at the previous time t, and the predicted value of the sample price data combination is a price data combination at the future predicted time corresponding to the previous time t, both of which are historical data.

[0025] Specifically, historical price data combinations are obtained on a daily basis. Each data combination includes the electricity price for each hour of the day, the carbon price for the day, and the green certificate price for the day. That is, a set of data includes 26 data points, and the data from each quarter for many years constitute the training data subset; the data from all quarters for many years constitute the training data set, that is, the training data set contains four training data subsets.

[0026] Step S3: determining a deviation vector of price data based on the combination of the historical price data and the predicted price.

[0027] In step S4, a multi-core-based single-class support vector machine algorithm is used to generate an uncertainty set of the predicted price based on the deviation vector of the price data.

[0028] In step S5, a neural network classifier is used to classify the uncertainty set of the predicted price to obtain multiple price scenarios; each price scenario includes multiple predicted prices.

[0029] Step S6, using the Pied Kingfisher optimization algorithm, based on the predicted prices under multiple price scenarios, solve the joint clearing model to obtain the operation strategy of the wind, solar and storage station; the joint clearing model includes a wind, solar and storage station power generation strategy model, a thermal power unit power generation strategy model, a first objective function and a first constraint condition; the wind, solar and storage station power generation strategy model includes a second objective function and a second constraint condition; the thermal power unit power generation strategy model includes a third objective function and a third constraint condition.

[0030] Step S7: regulating the wind-solar-storage station based on the operation strategy of the wind-solar-storage station.

[0031] As an optional implementation, in step S2, the LSTM neural network model based on the hybrid attention mechanism includes a convolutional neural network, a sequential convolutional attention module, and a long short-term memory recursive neural network connected in sequence.

[0032] Specifically, the LSTM neural network model structure based on the hybrid attention mechanism includes: a convolutional neural network (CNN), a sequential convolution attention module (SCAM), and a long short term memory recurrent neural network (LSTM). The input is a combination of historical price data at time t before the prediction moment, and the output is the predicted value of electricity price, carbon price, and green certificate price at the prediction moment. Various types of prices are typical chaotic time series with significant time dependence. As an excellent time feature extractor, LSTM can more comprehensively describe the changes in time series; in order to further improve the feature extraction ability of the model, CNN is introduced to extract features between data; considering a large number of different input features, a sequential convolution attention module SCAM is used to fuse feature maps between different data.

[0033] At the CNN layer, a CNN is used for unified feature extraction. The input is a combination of price data from the previous t time periods, and the output is a high-dimensional feature map. A ReLU activation function is applied between each convolutional layer and the pooling layer to ignore irrelevant features and improve model convergence speed. The pooling layer uses the max pooling method. The convolution process is shown in the following equation.

[0034]

[0035] Among them, I0 and O are the original input; I ai represents the feature output of the ai layer; f ReLU is the ReLU activation function; is the convolution operation; W ai is the weight vector of the convolution kernel of the ai-th layer; b ai is the bias vector of the ai-th layer; Q ai is the feature output of the ai-th pooling layer, and max represents the maximum pooling operation.

[0036] At the SCAM layer: For one-dimensional sequence prediction tasks that require more precise numerical information, a hybrid attention model, the Sequential Convolutional Attention Module (SCAM), is proposed, consisting of a channel attention model (CAM) and a temporal attention module (TAM). The CAM assigns weights to different feature maps, automatically selecting the most meaningful input features for each task. The TAM concatenates the input features selected by the CAM according to the temporal dimension, extracts convolution kernel features, and then uses the Sigmoid function to obtain the temporal attention matrix, achieving weight assignment in the temporal dimension. The SCAM layer inputs a high-dimensional feature map that captures important features of the sequence from both the channel and time dimensions, thus avoiding the loss of key information. The final output considers features of both the channel and time series.

[0037] In the LSTM layer: LSTM is a recurrent neural network used to process sequences with specific temporal dependencies. It effectively captures hidden temporal features. An LSTM unit consists of three gate structures: input gate, forget gate, and output gate. These gate structures work together to address the problems of vanishing and exploding gradients during training. The calculation method is as follows.

[0038]

[0039] Among them, ρ represents the LSTM related operation function, w t 、g t 、o t Represent the output features of the input gate, forget gate and output gate at the tth period, W w 、W g 、W o is the weight index, b w 、b g 、b o is the bias term, x t 、h t-1 are the input of the tth period and the output of the t-1th period respectively.

[0040] As an optional implementation, in step S2, the training process of the price prediction model specifically includes the following steps.

[0041] Step S21: Obtain a training data set.

[0042] Step S22: input the training data set into the LSTM neural network model based on the hybrid attention mechanism, and output the predicted value of the sample price data combination.

[0043] Step S23, construct a loss function based on the predicted value of the sample price data combination and the true value of the sample price data combination, and iteratively optimize the model parameters of the LSTM neural network model based on the hybrid attention mechanism according to the loss function until the number of iterations reaches a maximum value or the loss function reaches a minimum value, stop the iterative optimization, and obtain the price prediction model.

[0044] The training dataset is fed into a four-channel CNN-SCAM-LSTM network model to complete the training. After training, the CNN-SCAM-LSTM network model is fed with a combination of historical price data from the time period t before the prediction moment, and outputs the predicted price at the prediction moment.

[0045] As an optional implementation, step S3 specifically includes the following steps.

[0046] Step S31: Obtain an uncertainty dataset. The uncertainty dataset includes multiple data sets. Each data set is in the same format as the training dataset, except that they are datasets from different times. For example, if the training dataset is data from 2022, the uncertainty dataset can be data from 2023.

[0047] In step S32, multiple sets of data in the uncertainty data set are input into the CNN-SCAM-LSTM neural network model respectively to obtain the predicted price vector, which is compared with the true value (i.e., the corresponding historical price data combination) to obtain the deviation vector of the price data. The calculation formula is as follows.

[0048]

[0049] in, is the deviation vector of the price data of the nth group i; is the predicted price of the price data No. i in the nth group; It is the nth group and the i-th combination of historical price data.

[0050] As an optional implementation, in step S4, the uncertainty set of the predicted price is expressed as follows.

[0051]

[0052] Where U(D) is the uncertainty set of the predicted price; v is the uncertainty vector of the predicted price; α i * is the optimized value of the i-th dual variable in α; ε m * is the optimized value of the mth kernel function coefficient; K m (v,v i ) is the uncertainty vector v of the predicted price and the sample v of the mth kernel function iThe value between; p* is the optimized value of the bias term, where v i Represents the i-th deviation vector (i.e., the i-th sample) in the uncertainty vector of the predicted price.

[0053] Specifically, the deviation vector of price data is taken as input data, and multiple kernel functions are adopted. A method for generating the uncertainty set of predicted prices is proposed considering a one-class support vector machine (OC-SVM) based on multiple kernels (MKL). The expression of the OC-SVM based on MKL is as follows.

[0054]

[0055] Among them, M represents the number of basic kernel functions; H m is the coefficient vector of the support vector; ε m is the coefficient of each kernel function; p is the bias term; χ i is a slack variable that allows some points to deviate from the separating hyperplane; u o and μ o are regularization parameters used to control the degree of relaxation of the model; N is the number of samples; Represents the feature mapping function from the input space to the high-dimensional feature space.

[0056] We further use strong duality theory to obtain its dual problem. Assuming that all kernel functions are positive semidefinite, the dual problem is also convex and can be solved using a general convex optimization solver.

[0057]

[0058] Among them, α, γ, and λ are the dual variables corresponding to the constraints of the original problem, and K m (v i ,v j ) is the mth kernel function in sample v i and sample v j Values between α i and α j are the i-th and j-th dual variables in α respectively.

[0059] The solutions of the original problem (Formula (5)) and the dual problem (Formula (6)) are and {α * ,γ * ,λ *}; Using the solution result, its uncertainty set is expressed as formula (4).

[0060] As an optional implementation, step S5 specifically includes the following steps.

[0061] In step S51 , based on the predicted price uncertainty set in step S4 , Monte Carlo sampling is used to obtain a sample set as input data for a neural network classifier.

[0062] In step S52, a neural network classifier using particle swarm optimization is used to classify the sample set, generating multiple price scenarios. Since wind, solar, and storage stations formulate operational strategies (determine output) based on price data, different price data leads to different strategies. Due to the uncertainty and diversity of predicted prices, multiple price scenarios are used to represent them, i.e., each price scenario represents a type of predicted price data.

[0063] The price scenario is the arithmetic mean of the price data samples. Based on the classification results, multiple price scenarios can be obtained. First, the sample subclass of the predicted price data is calculated based on the deviation value of the predicted price at the prediction moment and the sample subclass. Then, the arithmetic mean of the sample subclass of the predicted price data is taken to obtain the price scenario under the sample subclass, as shown in Formula (7). The price scenario weight is defined as shown in Formula (8).

[0064]

[0065] in, is the deviation value of the price data of the nth group i under the xth price scenario; is the average predicted value of the price data No. i under the x-th price scenario, is the predicted price value of the i-th price data; N x is the number of data sets under the x-th price scenario; ξ x is the weight of the x-th price scenario.

[0066] Furthermore, the specific process of constructing a neural network classifier considering particle swarm optimization is as follows.

[0067] A neural network classifier consists of three parts: an input layer, a hidden layer, and an output layer. In the hidden layer, Fuzzy C-Means (FCM) clustering is used to calculate fitness values, and Least Squares Estimation (LSE) is used to calculate connection weights. In the output layer, fuzzy inference is used to calculate the classifier's output. Fuzzy inference is the process of mapping given input data to output data using membership functions, logical operators, and if-then rules.

[0068] In the premise of fuzzy rules, FCM clustering is used to group the data. FCM clustering is used in data analysis to find the centroid and degree of belonging of each cluster. The FCM clustering algorithm consists of the following steps.

[0069] 1) Determine the number of clusters and initialize the membership matrix U r .

[0070]

[0071] Among them, u ci,cj Represents sample x cj The membership degree belongs to cluster ci, CI is the total number of clusters, and CJ is the total number of samples.

[0072] 2) Calculate the centroid of each cluster.

[0073]

[0074] Among them, cv ci is the centroid of cluster ci, and FC represents the fuzzification coefficient.

[0075] 3) Update the membership matrix using the Euclidean distance function.

[0076]

[0077] Among them, ||x cj -v ci || is x cj to v ci The Euclidean distance of d ci,cj are the Euclidean distances from the sample to the cluster centroid; U r+1 is the membership matrix of the r+1th iteration.

[0078] 4) Check the termination condition. If the following conditions are met, terminate the iteration; otherwise, return to step 2 and continue the iteration.

[0079] ||U r+1 -U r ||≤δ u (12)

[0080] Among them, δ u It is a preset threshold used to determine whether the algorithm has converged.

[0081] In the inference phase of the fuzzy rules, the connection weight coefficients are estimated using the least squares method. LSE is a global learning algorithm that minimizes the overall squared error between the model output and the target output. The mean squared error (MSE) is used as the objective function of LSE.

[0082]

[0083] Among them, f ci (x cj ) is x cjThe output value of the next cluster ci; is the kth power of the cj-th output; KL is a constant; A k is the connection weight coefficient; X is the input data matrix; Y is the target output matrix.

[0084] The resulting value of the connection weight coefficient is determined by the following formula.

[0085] A k =(X T X) -1 X T Y k (14)

[0086] The particle swarm optimization (PSO) algorithm is further used to optimize the parameters of the neural network classifier. The parameters optimized by PSO are related to three factors: the number of fuzzy rules used in fuzzy c-means clustering, the value of the fuzzification coefficient, and the polynomial type of the connection weight. The process of using particle swarm optimization for neural network classifier parameters is as follows.

[0087] 1) Randomly generate a particle swarm Ja, particle position pa, and particle velocity va.

[0088] Ja(g)=[va1(g),va2(g),…,va b (g)] T (15)

[0089] Among them, b represents the number of parameters that need to be optimized, that is, the number of particles; g represents the number of times the current PSO iteration is executed.

[0090] 2) Adjust the inertia weight ra.

[0091]

[0092] Among them, ra max and ra min Represent the maximum and minimum values of inertia weight respectively; g max represents the maximum number of iterations; ra(g) is the inertia weight at the g-th iteration.

[0093] 3) Update particles: by using pa bt (g) and qa bt (g) value to adjust the particle's velocity va.

[0094] va(g+1)=ra(g)va(g)+c1e1[pa bt (g)-pa(g)]+c2e2[qa bt (g)-pa(g)](17)

[0095] Where va(g) and va(g+1) are the particle velocities at the g-th iteration and the g+1-th iteration, respectively; c1 and c2 are acceleration constants, e1 and e2 are random values between 0 and 1; pa bt (g) is the optimal position of the particle at the g-th iteration; qa bt (g) is the global optimal position at the g-th iteration; pa(g) is the attractor of the particle at the g-th iteration, which indicates the ability of the particle to attract other particles to approach.

[0096] 4) Update the position of each particle using the updated velocity, evaluate the updated particles using the objective function, and compare their position in pa bt (g) and qa bt (g) performance.

[0097] pa(g+1)=pa(g)+va(g+1) (18)

[0098] Among them, pa(g+1) is the attractor of the particle at the g+1th iteration.

[0099] 5) If the termination condition is not met, repeat steps 2 to 4.

[0100] The above process is explained as follows. The particle structure is the parameter that needs to be optimized; the parameters are determined by the particles, and the selectable parameters are as follows.

[0101]

[0102] Among them, z1 is the number of fuzzy rules; z2 is the value of FC fuzzification coefficient; z3 is the polynomial type of connection weight; Constant represents a constant; Linear represents a linear expression.

[0103] According to the particle swarm optimization results, the neural network classifier is constructed using the selected parameters.

[0104] As an optional implementation, in step S6, a joint clearing model is established based on the power generation strategies of the wind, solar and storage stations and the thermal power units. The corresponding objective function is the first objective function, and the expression of the first objective function is as follows.

[0105]

[0106] in, The price quote for the kth segment of the wind, solar and storage station at time t; is the output active power of the wind-solar-storage station at the kth segment at the tth moment; The price for the kth section of the thermal power unit at time t; is the output active power of the kth section of the thermal power unit at the tth moment.

[0107] In the joint clearing model, in addition to considering the power generation strategy constraints of wind, solar, and storage stations and thermal power units, it is required to meet the power load demand. Therefore, the power balance constraint is considered. The corresponding first constraint condition is as follows.

[0108]

[0109] Among them, p g,t is the actual output of the thermal power unit at time t; p w,t is the output active power of the wind turbine at the tth moment; p pv,t is the output active power of the photovoltaic unit at the tth moment; is the discharge power of the electrochemical energy storage system; Charging power for electrochemical energy storage system; p l,t The electrical load demand.

[0110] Among them, when establishing the wind-solar-storage station power generation strategy model, based on the predicted prices under multiple price scenarios, the wind-solar-storage station power generation strategy model formulates a power generation strategy with the goal of maximizing its own profits. The corresponding objective function is the second objective function, and the expression of the second objective function is as follows.

[0111]

[0112] Among them, ξ x is the weight of the x-th price scenario; is the electricity price at time t under the x-th price scenario; is the price of green certificates under the xth price scenario; a w,t is the green certificate allocation coefficient of the wind turbine; a pv,t It is the green certificate allocation coefficient of the photovoltaic unit.

[0113] The second constraint corresponding to the wind-solar-storage station power generation constraint is as follows.

[0114]

[0115]

[0116] in, is the predicted output of the photovoltaic unit at time t; is the installed capacity of the PV unit at time t; is the predicted output of the wind turbine at time t; is the installed capacity of the wind turbine at the tth moment; p ess,min is the minimum charge and discharge power of the electrochemical energy storage system; p ess,max SoC is the maximum charge and discharge power of the electrochemical energy storage system; ess,tState of charge of electrochemical energy storage system; SoC ess,min SoC is the lower limit of the state of charge of the electrochemical energy storage system; ess,max is the upper limit of the electrochemical energy storage system and state of charge; η ess,in is the charging efficiency of the electrochemical energy storage system; η ess,out is the discharge efficiency of the electrochemical energy storage system; is the rated energy of the electrochemical energy storage system.

[0117] When establishing a thermal power generation strategy model, the power generation strategy of the thermal power unit is formulated based on the predicted prices under multiple price scenarios and with the cost and carbon emission factors of the thermal power unit as the benchmark. The corresponding objective function is the third objective function, and the expression of the third objective function is as follows.

[0118]

[0119] in, is the carbon sink price under the xth price scenario; X zi is the initial carbon quota within the scheduling period; b g,t is the total carbon emission factor of the thermal power unit at time t; b coal,t is the carbon emission factor of coal consumption of thermal power unit at time t; b loss,t is the loss carbon emission factor of the thermal power unit at time t; b oil,t is the carbon emission factor of the thermal power unit at the time t; It is the lower limit output of thermal power units during the normal peak regulation stage; It is the upper limit of the output of thermal power units during the normal peak regulation stage; It is the lower limit output of thermal power units during the speed reduction and peak regulation stage; It is the lower limit output of thermal power units during the oil-injection peak-shaving stage.

[0120] Among them, the deep peak regulation model of thermal power units is as follows.

[0121]

[0122] Among them, C g,t is the total cost of the thermal power unit at time t; C coal,t is the coal consumption cost of the thermal power unit at time t; C loss,t is the loss cost of the thermal power unit at time t; C oil,t is the oil input cost of the thermal power unit at time t; p g,t is the actual output of the thermal power unit at moment t.

[0123] Introducing binary variable linearization segmentation constraints, the total cost C of thermal power units g,t It is expressed as follows.

[0124]

[0125] in, and are binary variables representing the oil injection peak regulation phase and speed reduction peak regulation phase respectively.

[0126] The unit's consumption characteristics and energy cost characteristics can be expressed as piecewise functions divided by the peak-shaving stage. The unit's carbon emissions and coal consumption have a strong positive correlation and can also be divided according to the unit's peak-shaving stage. The specific expression is shown in Equation (26).

[0127] The deep peak regulation of thermal power units must meet the ramping constraint and technical output constraint, which are taken as the third constraint condition. The third constraint condition is as follows.

[0128]

[0129] in, The minimum technical output of thermal power units; The maximum technical output of the thermal power unit; It is the upward climbing power limit of the thermal power unit; It is the downward power limit of thermal power units.

[0130] As an optional implementation, in step S6, the Pied Kingfisher optimization algorithm is used to solve the joint clearing model that considers the above wind, solar and storage stations and the power generation strategies of thermal power units. Finally, after the solution, the operation strategy of the wind, solar and storage stations (i.e., the output of the wind, solar and storage stations at each moment) can be obtained.

[0131] The Pied Kingfisher Optimization Algorithm combines three foraging operations to simulate the foraging strategy of the Pied Kingfisher: the perching and circling phase, the diving phase, and the symbiotic phase. This algorithm effectively solves various optimization challenges in various search spaces and achieves computational accuracy requirements at a low computational cost. The three PKO strategies are shown below, and the algorithm optimization process is as follows.

[0132] First, the population is initialized and a set of initial solutions are randomly generated from the search space as the first trial to start the search process. Second, random numbers are generated and the Pied Kingfisher will choose between the perching and circling phases, i.e., Equation (32); and the diving phase, i.e., Equation (33), to update the position of the model solution. Furthermore, the Pied Kingfisher enters the symbiotic phase, i.e., Equation (34). After the position is updated, the updated position X is evaluated. n The target fitness value at (t+1) is better than the current best position X. best (t), then the best position is replaced; finally, it continues to iterate until the set maximum number of iterations is reached, and the best position result, that is, the optimal solution, is output.

[0133] X n (t+1)=Xn (t)+α pko L pko ×(X m (t)-X n (t)),m≠n,m,n∈{1,2,...,N pko}(32)

[0134] X n (t+1)=X n (t)+DS*o pko *α pko *(b pko -X best (t)),n∈{1,2,...,N pko}(33)

[0135]

[0136] Among them, X n (t) and X n (t+1) is the position of individual n at the tth iteration and the t+1th iteration, X m (t) is the position of individual m at the tth iteration; α pko is a random parameter; N pko is the population size; L pko are dynamic parameters related to perching and circling behaviors; pko and DS are the control parameters representing hunting ability, and pko and X best (t) related position parameters; X r1 (t) and X r2 (t) is two individuals randomly selected from the population; PE is the predation efficiency parameter; rand is a random number between 0 and 1.

[0137] The beneficial effects of this application are as follows.

[0138] 1) A neural network model integrating CNN-SCAM-LSTM was established, which has excellent temporal feature extraction capabilities, can comprehensively describe the changes in time series, effectively capture the long-term and short-term dependencies between electricity prices, carbon sink prices and green certificate prices in time series data, and improve the prediction accuracy of price prediction tasks.

[0139] (2) By using multiple kernel functions to adapt to the distribution and characteristics of different prediction data, a prediction price uncertainty set based on a multi-kernel single-class support vector machine is proposed to effectively handle outliers and noise data and improve the reliability of the uncertainty set.

[0140] (3) For the neural network classifier for data set classification, the particle swarm optimization algorithm is used to optimize the parameters of the neural network, reduce the neural network training time, improve efficiency, and significantly improve the accuracy and stability of price scenario classification.

[0141] (4) Establishing a power generation strategy model for wind, solar and storage stations under various price scenarios can achieve benefit stability under price uncertainty, avoid excessive losses, and have excellent robustness. At the same time, the use of the Pied Kingfisher optimization algorithm to solve the joint clearing model can effectively solve various optimization problems in various search spaces and achieve computational accuracy requirements at a lower computational cost.

[0142] In an exemplary embodiment, a computer device is provided, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a wind, solar, and storage station control method that considers electricity, green, and carbon price forecasts.

[0143] In an exemplary embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, a wind-solar-storage station control method considering electricity-green-carbon price forecast is implemented.

[0144] In an exemplary embodiment, a computer program product is provided, comprising a computer program that, when executed by a processor, implements a method for controlling a wind, solar, and storage station in consideration of electricity, green, and carbon price forecasts.

[0145] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it realizes a wind-solar storage station control method taking into account the electricity-green-carbon price forecast.

[0146] Those skilled in the art will understand that Figure 3The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0147] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0148] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0149] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0150] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0151] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A wind-solar-storage station control method considering electricity-green-carbon price forecast, characterized in that: The wind-solar-storage station control method considering electricity-green-carbon price forecasts includes: Obtain a combination of historical price data; the combination of historical price data includes: historical carbon prices, historical green certificate prices, and historical electricity prices; The historical price data combination is input into a price prediction model to output a predicted price; the predicted price includes: a predicted carbon price value, a predicted green certificate price value, and a predicted electricity price value; the price prediction model is obtained by training an LSTM neural network model based on a hybrid attention mechanism using a training data set; the training data set includes a sample price data combination and the corresponding predicted value of the sample price data combination; determining a deviation vector of price data based on the combination of the historical price data and the predicted price; A multi-core based single-class support vector machine algorithm is used to generate an uncertainty set for the predicted price based on the deviation vector of the price data; A neural network classifier is used to classify the uncertainty set of the predicted price to obtain multiple price scenarios; each price scenario includes multiple predicted prices; Using the Pied Kingfisher optimization algorithm, the joint clearing model is solved based on the predicted prices under multiple price scenarios to obtain the operation strategy of the wind, solar and storage station. The joint clearing model includes a wind, solar and storage station power generation strategy model, a thermal power unit power generation strategy model, a first objective function, and a first constraint condition. The wind, solar and storage station power generation strategy model includes a second objective function and a second constraint condition. The thermal power unit power generation strategy model includes a third objective function and a third constraint condition. Regulate the wind-solar-storage station based on its operation strategy; The expression of the first objective function is: in, The price quote for the kth segment of the wind, solar and storage station at time t; is the output active power of the wind-solar-storage station at the kth segment at the tth moment; The price for the kth section of the thermal power unit at time t; is the output active power of the kth section of the thermal power unit at time t; The first constraint is: Among them, p g,t is the actual output of the thermal power unit at time t; p w,t is the output active power of the wind turbine at the tth moment; p pv,t is the output active power of the photovoltaic unit at the tth moment; is the discharge power of the electrochemical energy storage system; Charging power for electrochemical energy storage system; p l,t For electrical load demand.

2. The wind-solar-storage station control method considering electricity-green-carbon price forecast according to claim 1 is characterized in that: The LSTM neural network model based on the hybrid attention mechanism includes a sequentially connected convolutional neural network, a sequential convolutional attention module, and a long short-term memory recurrent neural network.

3. The wind-solar-storage station control method considering electricity-green-carbon price forecast according to claim 2 is characterized in that: The training process of the price prediction model specifically includes: Get the training dataset; Input the training data set into an LSTM neural network model based on a hybrid attention mechanism, and output the predicted value of the sample price data combination; A loss function is constructed based on the predicted value of the sample price data combination and the true value of the sample price data combination, and the model parameters of the LSTM neural network model based on the hybrid attention mechanism are iteratively optimized according to the loss function until the number of iterations reaches the maximum value or the loss function reaches the minimum value. The iterative optimization is stopped to obtain a price prediction model.

4. The wind-solar-storage station control method considering electricity-green-carbon price forecast according to claim 1 is characterized in that: The uncertainty set of the predicted price is expressed as: Where U(D) is the uncertainty set of the predicted price; v is the uncertainty vector of the predicted price; α i * is the optimized value of the i-th dual variable in α; ε m * is the optimized value of the mth kernel function coefficient; K m (v,v i ) is the uncertainty vector v of the m-th kernel function in predicting the price and the sample v i ; p* is the optimized value of the bias term.

5. The wind-solar-storage station control method considering electricity-green-carbon price forecast according to claim 1 is characterized in that: The expression of the second objective function is: Among them, ξ x is the weight of the x-th price scenario; is the electricity price at time t under the x-th price scenario; is the price of green certificates under the xth price scenario; a w,t is the green certificate allocation coefficient of the wind turbine; a pv,t Green certificate allocation coefficient for photovoltaic units; The second constraint is: in, is the predicted output of the photovoltaic unit at time t; is the installed capacity of the PV unit at time t; is the predicted output of the wind turbine at time t; is the installed capacity of the wind turbine at time t; p ess,min is the minimum charge and discharge power of the electrochemical energy storage system; p ess,max SoC is the maximum charge and discharge power of the electrochemical energy storage system; ess,t State of charge of electrochemical energy storage system; SoC ess,min SoC is the lower limit of the state of charge of the electrochemical energy storage system; ess,max is the upper limit of the electrochemical energy storage system and state of charge; η ess,in is the charging efficiency of the electrochemical energy storage system; η ess,out is the discharge efficiency of the electrochemical energy storage system; is the rated energy of the electrochemical energy storage system.

6. The wind-solar-storage station control method considering electricity-green-carbon price forecast according to claim 5 is characterized in that: The expression of the third objective function is: in, is the carbon sink price under the xth price scenario; X zi is the initial carbon quota in the scheduling period; b g,t is the total carbon emission factor of the thermal power unit at time t; b coal,t is the carbon emission factor of coal consumption of thermal power unit at time t; b loss,t is the loss carbon emission factor of the thermal power unit at time t; b oil,t is the carbon emission factor of the thermal power unit at the time t; It is the lower limit output of thermal power units during the normal peak regulation stage; It is the upper limit of the output of thermal power units during the normal peak regulation stage; It is the lower limit output of thermal power units during the speed reduction and peak regulation stage; It is the lower limit output of the thermal power unit during the oil-input peak-shaving stage; The third constraint is: in, The minimum technical output of thermal power units; The maximum technical output of the thermal power unit; It is the upward climbing power limit of the thermal power unit; It is the downward ramp power limit of thermal power units.

7. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the wind-solar-storage station control method considering electricity-green-carbon price forecasts as described in any one of claims 1-6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the wind, solar and storage station control method considering electricity, green and carbon price prediction according to any one of claims 1 to 6 is implemented.

9. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the wind, solar and storage station control method considering electricity, green and carbon price prediction according to any one of claims 1 to 6 is implemented.

Images