Micro-grid optimization scheduling method and system

Abstract

The invention relates to a micro-grid optimization scheduling method and system, and belongs to the technical field of power grid scheduling. The method comprises the following steps: establishing a micro-grid optimization scheduling model which takes the minimum new energy abandoned energy in a certain typical scene or the minimum new energy abandoned energy sum in two or more typical scenes in a prediction interval of prediction quantity as a target and takes energy storage charging and discharging power as a decision variable, and solving the model to obtain a micro-grid optimization scheduling plan. According to the invention, the uncertainty of the uncertainty factors during the operation of the micro-grid is quantified through interval prediction, and the micro-grid optimization scheduling model considering the uncertainty of the uncertainty factors is established and solved. The solved microgrid optimization scheduling plan can guarantee the consumption level of new energy and the operation efficiency of the system under the uncertain situation of the uncertain factors in the prediction interval, and the operation stability of the microgrid is improved. According to the method, quantification of uncertainty relates to a mechanism level, and compared with existing micro-grid optimization scheduling, the method is more refined.

Description

Technical Field

[0001] The present invention relates to a microgrid optimal scheduling method and system, belonging to the technical field of power grid scheduling. Background Art

[0002] As a small power system for balancing the access of new energy and the safe and stable operation of the large power grid, reasonably coping with the uncertainty of new energy output is one of the key links to ensure its stable and efficient operation. In terms of the optimal scheduling of microgrids, current technologies mainly formulate scheduling plans based on the deterministic prediction results of new energy output or select typical scenarios such as typical days / weeks from historical observation data, highly relying on the accuracy of new energy prediction, and insufficiently considering the uncertainty characteristics of new energy output (such as the output of wind and solar power plants) and load. During actual operation, this may lead to problems such as a decrease in the renewable energy consumption rate, a decrease in system operation efficiency, and an increase in the risks of power grid stability and security.

[0003] The patent document with publication number CN116760121A discloses a low-carbon scheduling method, system, device and medium for a virtual power plant. The scheduling method specifically includes the following steps: inputting the power data of the current time period into a power prediction model to predict the power data of the future time period, where the power data includes the output power of wind turbines, the output power of photovoltaic units, and the user load power; constructing constraint conditions containing fuzzy parameters, where the constraint conditions include power balance constraints and spinning reserve constraints; the fuzzy parameters characterize the uncertainty of wind turbines, photovoltaic units, and user loads; at a set confidence level, using a clear equivalence class to process the constraint conditions containing fuzzy parameters to obtain deterministic constraint conditions; constructing an optimal scheduling model for the virtual power plant; and solving the objective function according to the power data of the future time period and the scheduling constraint conditions to obtain the optimal power data of the future time period. This scheduling method considers the uncertainty of wind / solar and load, but this scheduling method incorporates uncertainty into the constraint conditions of the scheduling stage. Essentially, it copes with uncertainty by providing redundant reserves, lacks precise quantification of uncertainty, and this method is prone to excessive redundancy, thus reducing the operation efficiency of the system.

[0004] The patent document with the publication number CN114662798A discloses a dispatching method, device and electronic equipment based on the economic operation domain of the power grid. Aiming at the problem that the traditional dispatching plan formulation is conservative due to the uncertainty factors of the power grid, this dispatching method proposes an economic dispatching method that can quantitatively analyze the impact of the uncertainty of uncertainty factors on the power grid dispatching plan arrangement: determining the uncertainty factors and dispatching objectives that need to be considered when formulating the power grid dispatching plan; rolling and depicting the economic operation domain of the power grid according to the uncertainty factors and dispatching objectives, and selecting the form of the economic operation domain of the power grid according to the actual engineering needs; calculating the optimal dispatching plan in real time based on the economic operation domain of the power grid. Among them, the uncertainty factors include the uncertainty of new energy output, the uncertainty of load level, the uncertainty of unplanned outage of equipment, the uncertainty of natural disasters and the uncertainty of human factor risks faced by the power grid during operation. This method considers the uncertainty factors of the power grid, specifically by establishing a set of uncertainty factors, but does not involve the precise quantification of the uncertainty of each factor. Therefore, problems such as low operation stability of the microgrid and low new energy consumption level are still likely to occur due to inaccurate uncertainty quantification and low accuracy. Summary of the Invention

[0005] The object of the present invention is to provide a microgrid optimal dispatching method and system different from the existing ones to solve the problems of low operation stability of the microgrid and low new energy consumption level caused by the low accuracy of uncertainty quantification of uncertainty factors.

[0006] To achieve the above object, the solution of the present invention includes:

[0007] A microgrid optimal dispatching method of the present invention includes the following steps:

[0008] Establish a microgrid optimal dispatching model with the minimum new energy curtailment in a certain typical scenario within the prediction interval of the predicted value or the minimum sum of new energy curtailment in two or more typical scenarios as the objective and the charge and discharge power of energy storage as the decision variable, and solve this model to obtain the microgrid optimal dispatching plan;

[0009] When the condition of ignoring the sequence noise of the predicted value is satisfied, the prediction interval is taken as:

[0010]

[0011] Otherwise it is taken as:

[0012]

[0013] Among them, and Respectively corresponding to the predicted quantity: the estimated deterministic relationship, the variance of the sequence noise, and the variance of the prediction error. The estimated deterministic relationship is the estimated deterministic relationship between the true value of the predicted quantity and the corresponding predictor. The true value of the predicted quantity is equal to the sum of the estimated deterministic relationship corresponding to the predicted quantity, the sequence noise, and the prediction error. The sequence noise follows a normal distribution with a mean of 0 and a fixed variance of and the prediction error follows a normal distribution with a mean of 0 and a fixed variance of The predicted quantity is the uncertain factor faced during the operation of the microgrid. is the critical value of the standard normal distribution at the confidence level α.

[0014] Furthermore, the condition for ignoring the sequence noise is:

[0015]

[0016] or

[0017]

[0018] where 10 -q is the precision level of the critical value.

[0019] Furthermore, and are calculated based on the prediction dataset of the predicted quantity, the bootstrap theory, and the neural network model.

[0020] Furthermore, The calculation formula of is as follows:

[0021]

[0022] where represents the prediction result of the k-th training set, K is the number of elements in the set Ω, and the prediction results with a prediction accuracy higher than the threshold are selected from the m prediction results to form the set Ω.

[0023] Furthermore, The calculation formula of is as follows:

[0024]

[0025] where y i is the true value.

[0026] Furthermore, The calculation formula of is as follows:

[0027]

[0028] where Denote the prediction result of the k-th training set. K is the number of elements in the set Ω. Among the m groups of prediction results, the prediction results with a prediction accuracy higher than the threshold are selected to form the set Ω.

[0029] Furthermore, the uncertainty factors at least include new energy output and load.

[0030] Furthermore, the typical scenarios are combinations of new energy scenarios and load scenarios. The new energy scenarios include the maximum, minimum, and strongest fluctuation of new energy output, and the load scenarios include the maximum, minimum, and strongest fluctuation of load.

[0031] Furthermore, the strongest fluctuation of new energy output is that the adjacent new energy output prediction points respectively take the minimum and maximum values within their corresponding prediction intervals; the strongest fluctuation of load is that the adjacent load prediction points respectively take the minimum and maximum values within their corresponding prediction intervals.

[0032] A microgrid optimal scheduling system of the present invention includes a processor, and the processor is used to execute a computer program to implement the steps of the microgrid optimal scheduling method as described above.

[0033] The beneficial effects of the present invention:

[0034] The present invention is a pioneering invention, which provides a microgrid optimal scheduling method. By interval prediction, the uncertainty of the uncertainty factors faced during the operation of the microgrid is quantified, and a microgrid optimal scheduling model considering the uncertainty of the uncertainty factors is established and solved. The microgrid optimal scheduling model aims to minimize the new energy curtailment under a certain typical scenario within the prediction interval of the predicted quantity, with the energy storage charge and discharge power as the decision variable, or the microgrid optimal scheduling model aims to minimize the sum of new energy curtailment under two or three or more typical scenarios within the prediction interval of the predicted quantity, with the energy storage charge and discharge power as the decision variable. The obtained microgrid optimal scheduling plan, that is, the obtained microgrid optimal operation process, can ensure the new energy consumption level and system operation efficiency under the uncertainty scenarios of the uncertainty factors within the prediction interval, thereby improving the operation stability of the microgrid.

[0035] The microgrid optimal scheduling method of the present invention deeply analyzes the mechanism of uncertainty generation in the prediction stage and accurately quantifies the uncertainty of uncertainty factors based on statistical theory. In the scheduling stage, a processing strategy considering the uncertainty boundary (covering multiple typical scenarios) is proposed, which not only effectively copes with uncertainty but also takes into account the system operation efficiency, making the generated scheduling result (microgrid optimal scheduling plan) more scientific and efficient. Among them, the mechanism of uncertainty generation includes its own fluctuation (sequence noise) and prediction error. The accuracy of uncertainty quantification is the basic condition for the effective operation of the microgrid optimal scheduling method of the present invention, and the quantification of uncertainty in the microgrid optimal scheduling method of the present invention involves the mechanism level, which is more scientific and refined than the existing microgrid optimal scheduling.

[0036] Taking the uncertainty of new energy output and load faced by the microgrid during operation as an example, the uncertainty of new energy output and load of the microgrid is quantified through interval prediction, and a microgrid optimal scheduling model considering the double uncertainty of new energy output and load is established and solved. The microgrid optimal scheduling model takes the minimum new energy curtailment under a certain typical scenario within the prediction interval of the predicted value or the minimum sum of new energy curtailment under two or more typical scenarios as the objective, and the charge and discharge power of the energy storage as the decision variable. The obtained microgrid optimal scheduling plan, that is, the obtained microgrid optimal operation process, can ensure the consumption level of new energy and the system operation efficiency under the double uncertainty scenario of new energy output and load within the prediction interval, thereby improving the operation stability of the microgrid. Brief Description of the Drawings

[0037] Figure 1 is the flowchart of the microgrid optimal scheduling considering the double uncertainty of new energy output and load;

[0038] Figure 2 is the flowchart of the bootstrap-neural network uncertainty prediction. Detailed Implementation Manner

[0039] To solve the problems in the background technology, the present invention provides a microgrid optimal scheduling method, which quantifies the uncertainty of uncertainty factors faced by the microgrid during operation through interval prediction, establishes and solves a microgrid optimal scheduling model considering the uncertainty of uncertainty factors faced by the microgrid during operation. The obtained microgrid optimal operation process can ensure the consumption level of new energy and the system operation efficiency under the uncertainty scenario of uncertainty factors within the prediction interval, thereby improving the operation stability of the microgrid. The quantification of uncertainty in the present invention involves the mechanism level, which is more scientific and refined than the existing microgrid optimal scheduling.

[0040] To make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0041] An embodiment of a microgrid optimal scheduling method:

[0042] A microgrid optimal scheduling method includes the following steps:

[0043] Establish a microgrid optimal scheduling model with the objective of minimizing the curtailment of new energy in a certain typical scenario within the prediction interval of the predicted quantity or minimizing the sum of the curtailment of new energy in two or more typical scenarios, and using the charge and discharge power of energy storage as the decision variable. Solve this model to obtain the microgrid optimal scheduling plan. The microgrid optimal scheduling plan refers to the microgrid optimal operation process, through which the consumption level of new energy and the system operation efficiency can be guaranteed under the uncertain scenarios of the uncertain factors faced by the microgrid during operation within the prediction interval, thereby improving the operation stability of the microgrid.

[0044] Among them, when the condition of ignoring the sequence noise of the predicted quantity holds, only consider the prediction error of the predicted quantity. The prediction interval of the predicted quantity is specifically taken as:

[0045]

[0046] Otherwise, consider the sequence noise and prediction error of the predicted quantity. The prediction interval of the predicted quantity is specifically taken as:

[0047]

[0048] Among them, and are respectively the corresponding: estimated deterministic relationship, variance of sequence noise, and variance of prediction error of the predicted quantity; the estimated deterministic relationship is the estimated deterministic relationship between the true value of the predicted quantity and the corresponding prediction factor; the true value of the predicted quantity is equal to the sum of the corresponding: estimated deterministic relationship, sequence noise, and prediction error; the sequence noise follows a normal distribution with a mean of 0 and a fixed variance of the variance of the sequence noise The mathematical expression of the sequence noise is sequence noise The prediction error follows a normal distribution with a mean of 0 and a fixed variance of the variance of the prediction error The mathematical expression of the prediction error is prediction error The predicted quantity is the uncertain factor faced by the microgrid during operation; is the critical value of the standard normal distribution at the confidence level α.

[0049] Specifically, the condition for ignoring the sequence noise is:

[0050]

[0051] or

[0052]

[0053] wherein, 10 -q is the precision level of the critical value. Of course, the ignoring condition of the sequence noise can also be determined according to experience.

[0054] Specifically, and are calculated based on the prediction data set based on the predicted value, the bootstrap theory, and the neural network model. Of course, and can also be determined according to experience.

[0055] Wherein, The calculation formula is as follows:

[0056]

[0057] wherein, represents the prediction result of the k-th group of bootstrap training sets, K is the number of elements in the set Ω, and the prediction results with a prediction accuracy higher than the threshold are selected from the m groups of prediction results to form the set Ω.

[0058] Wherein, The calculation formula is as follows:

[0059]

[0060] wherein, y i is the true value.

[0061] Wherein, The calculation formula is as follows:

[0062]

[0063] wherein, represents the prediction result of the k-th group of bootstrap training sets, K is the number of elements in the set Ω, and the prediction results with a prediction accuracy higher than the threshold are selected from the m groups of prediction results to form the set Ω.

[0064] Specifically, the uncertainty factors faced during the operation of the microgrid include at least the new energy output and the load.

[0065] As other implementation manners, the uncertainty factors faced during the operation of the microgrid only include the new energy output, or may only include the load. Of course, it may also include risks such as unplanned outages of equipment, natural disasters, and human factors according to needs.

[0066] Taking the uncertainty factors faced by the microgrid during operation only including new energy output and load as an example, a microgrid optimal scheduling method includes the following steps:

[0067] Establish a microgrid optimal scheduling model with the goal of minimizing new energy curtailment in a certain typical scenario within the prediction interval of the predicted value or minimizing the sum of new energy curtailment in two or more typical scenarios, and using the energy storage charge and discharge power as the decision variable. Solve this model to obtain the microgrid optimal scheduling plan. The microgrid optimal scheduling plan refers to the microgrid optimal operation process, through which the consumption level of new energy and the system operation efficiency can be guaranteed under the dual uncertainty coupling scenario (dual uncertainty scenario) of new energy output and load within the prediction interval, thereby improving the operation stability of the microgrid.

[0068] Among them, when the condition of ignoring the sequence noise of new energy output holds, the prediction interval of new energy output is taken as:

[0069]

[0070] Otherwise, it is taken as:

[0071]

[0072] Among them, and are respectively the estimated deterministic relationship, the variance of the sequence noise, and the variance of the prediction error of the new energy output calculated based on the prediction data set of the new energy output, the bootstrap theory, and the neural network model; this estimated deterministic relationship is the estimated deterministic relationship between the true value of the new energy output and its prediction factors; the true value of the new energy output is equal to the sum of the estimated deterministic relationship of the new energy output, the sequence noise, and the prediction error; the sequence noise of the new energy output follows a normal distribution with a mean of 0 and a fixed variance of the variance of the sequence noise of the new energy output The mathematical expression is that the sequence noise of the new energy output The prediction error of the new energy output follows a normal distribution with a mean of 0 and a fixed variance of the variance of the prediction error of the new energy output The mathematical expression is that the prediction error of the new energy output is the critical value of the standard normal distribution at the confidence level α.

[0073] Among them, when the condition of ignoring the sequence noise of the load holds, the prediction interval of the load is taken as:

[0074]

[0075] Otherwise, it is taken as:

[0076]

[0077] Among them, and are the estimated deterministic relationship, the variance of the sequence noise, and the variance of the prediction error of the load calculated based on the load-based prediction data set, the bootstrap theory, and the neural network model, respectively; the estimated deterministic relationship is the estimated deterministic relationship between the true value of the load and its predictors; the true value of the load is equal to the sum of the estimated deterministic relationship of the load and the sequence noise and the prediction error; the sequence noise of the load follows a normal distribution with a mean of 0 and a fixed variance equal to the variance of the sequence noise of the load mathematically expressed as the sequence noise of the load The prediction error of the load follows a normal distribution with a mean of 0 and a fixed variance equal to the variance of the prediction error of the load mathematically expressed as the prediction error of the load is the critical value of the standard normal distribution at the confidence level α.

[0078] Among them, theoretically, at the confidence level α, the new energy output and the load may be any values within their respective prediction intervals, which are the uncertainties of the new energy output and the load respectively; the dual uncertainty coupling scenario of the new energy output and the load refers to considering both the uncertainty of the new energy output and the uncertainty of the load. There are theoretically countless scenarios. The typical scenarios of this solution are 16 relatively extreme scenarios selected from the dual uncertainty scenarios. The scheduling plans formulated under these extreme scenarios can, on the one hand, ensure the absorption rate under extreme scenarios, and on the other hand, reserve more adjustment space for real-time control under non-extreme scenarios.

[0079] Specifically, the typical scenarios are combinations of new energy scenarios and load scenarios. The new energy scenarios include the scenario of maximum new energy output, the scenario of minimum new energy output, and the scenario of the strongest new energy output fluctuation. The load scenarios include the scenario of maximum load, the scenario of minimum load, and the scenario of the strongest load fluctuation.

[0080] The existing typical scenarios are to select typical days or typical weeks or typical months, etc. with similar conditions from historical measured values for formulating scheduling plans, while the typical scenarios of the present invention are selected based on the interval prediction results. Different from the selection of existing typical scenarios: the existing deterministic prediction results can only obtain one typical scenario, while the typical scenarios (typical scenarios) of the present invention can extract countless scenarios from the interval prediction results; the historical typical scenarios are scenarios extracted or selected from historical data.

[0081] Among them, the scenario with the strongest new energy output fluctuation is that the adjacent new energy output prediction points respectively take the minimum and maximum values within their corresponding prediction intervals. Taking 20 new energy output prediction points as an example, the scenario with the strongest new energy output fluctuation is that the 1st new energy output prediction point in order takes the minimum value of its prediction interval, the 2nd new energy output prediction point takes the maximum value of its prediction interval, the 3rd new energy output prediction point takes the minimum value of its prediction interval, ……, the 20th new energy output prediction point takes the maximum value of its prediction interval; or the scenario with the strongest new energy output fluctuation is that the 1st new energy output prediction point in order takes the maximum value of its prediction interval, the 2nd new energy output prediction point takes the minimum value of its prediction interval, the 3rd new energy output prediction point takes the maximum value of its prediction interval, ……, the 20th new energy output prediction point takes the minimum value of its prediction interval. Of course, the scenario with the strongest new energy output fluctuation can also adopt other forms according to needs.

[0082] Among them, the scenario with the strongest load fluctuation is that the adjacent load prediction points respectively take the minimum and maximum values within their corresponding prediction intervals. Taking 10 load prediction points as an example, the scenario with the strongest load fluctuation is that the 1st load prediction point in order takes the minimum value of its prediction interval, the 2nd load prediction point takes the maximum value of its prediction interval, the 3rd load prediction point takes the minimum value of its prediction interval, ……, the 10th load prediction point takes the maximum value of its prediction interval; or the scenario with the strongest load fluctuation is that the 1st load prediction point in order takes the maximum value of its prediction interval, the 2nd load prediction point takes the minimum value of its prediction interval, the 3rd load prediction point takes the maximum value of its prediction interval, ……, the 10th load prediction point takes the minimum value of its prediction interval. Of course, the scenario with the strongest load fluctuation can also adopt other forms according to needs.

[0083] Among them, the prediction data set of the predicted quantity includes the prediction data set of the new energy output and the prediction data set of the load.

[0084] Specifically, the prediction data set of the new energy output is constructed based on the prediction factors of the new energy output.

[0085] Specifically, the prediction factors of the new energy output include meteorological data, and the meteorological data includes wind speed, temperature, light intensity and humidity.

[0086] Specifically, the prediction data set of the load is constructed based on the prediction factors of the load.

[0087] Specifically, the prediction factors of the load include meteorological data, holiday attributes and seasonal characteristics.

[0088] Specifically, the new energy includes wind power and photovoltaic power; as other implementation manners, the new energy only includes photovoltaic power or only includes wind power.

[0089] Specifically, an intelligent optimization algorithm is used to solve the microgrid optimal scheduling model, and the intelligent optimization algorithm includes the particle swarm optimization algorithm and the ant colony algorithm.

[0090] To ensure the stable operation of the microgrid and the operation of new energy and energy storage devices in a suitable state, the constraint conditions of the microgrid optimal scheduling model are set with power balance constraints, new energy output constraints, and energy storage state constraints.

[0091] Among them, the power balance constraint is:

[0092]

[0093] The new energy output constraint is:

[0094]

[0095] The energy storage state constraint is:

[0096]

[0097] In the formula, is the new energy curtailment power of the microgrid; and respectively represent the new energy output, load power, energy storage charging power (the energy storage charging power takes a negative value for discharging and a positive value for charging), and grid power supply power of the microgrid at time t in the typical scenario s; respectively represent the predicted power generation of the overall new energy, wind power, and photovoltaic power at time t in the typical scenario s; and respectively represent the maximum output power of wind power and photovoltaic power in the microgrid; and respectively represent the minimum and maximum allowable battery levels of the microgrid energy storage and the charging level at time t in the typical scenario s; Δt is the time interval from t - 1 to t; is the maximum value of the charging / discharging power of the microgrid energy storage, that is, in the charging mode, is the maximum value of the charging power of the microgrid energy storage, and in the discharging mode, is the maximum value of the discharging power of the microgrid energy storage.

[0098] The present invention divides the uncertainties of new - energy output and load into two parts: sequence noise and prediction error. Based on the bootstrap theory and artificial neural network, the two parts are quantified respectively. The specific idea is as follows: Based on the bootstrap theory, it is considered that both the sequence noise and the prediction error follow a normal distribution with a mean of 0. By continuously resampling with replacement, bootstrap subsets based on the observed sample set are constructed. Through each subset, a neural network model for the prediction target (predicted quantity) is established, and the error characteristics of the prediction for each subset are statistically analyzed to evaluate the prediction error of the prediction model. Since the prediction error follows a normal distribution with a mean of 0, as long as the number of subsets is large enough, the prediction error can be eliminated through the average integration of the prediction results. The difference between the prediction results of each subset and the average integrated value is the prediction error caused by the fact that the model cannot fully fit the relationship between the actual prediction factors and the prediction target. The difference between the prediction results of each subset and the measured value is the overall error including the prediction error and the sequence noise. In the training set and the test set, the overall error can be calculated from the measured value. Based on the properties of the normal distribution, the sequence noise can also be calculated from the statistical characteristics of the overall error and the statistical characteristics of the prediction error. By constructing a neural network model for the sequence noise of the prediction sample set, the sequence - noise characteristics of each future point to be predicted can be obtained, and the overall error characteristics (i.e., uncertainty) of each predicted point can be calculated by combining the prediction - error characteristics.

[0099] By quantifying the overall uncertainties of new - energy sources (such as wind and light) output and load, extreme scenarios of the micro - grid during the prediction period are obtained. With the goal of minimizing the curtailment of new energy in a certain extreme scenario, or minimizing the sum of new - energy curtailments in two extreme scenarios, or minimizing the sum of new - energy curtailments in three or more extreme scenarios, a micro - grid optimal scheduling model is established, and the solution results can ensure the new - energy consumption level of the micro - grid under uncertain scenarios (extreme scenarios).

[0100] A micro - grid optimal scheduling method, as Figure 1 and Figure 2 shown, the specific implementation steps are as follows.

[0101] Step 1: Obtain the historical output curves of new - energy sources such as wind and light, load curves, and meteorological data such as wind speed, temperature, light intensity, and humidity during the operation of the micro - grid. Respectively screen the prediction factors for the new - energy output, and construct a prediction data set D for the new - energy output. Among them, taking wind power and photovoltaic power as examples of new - energy output, that is, screen the prediction factors for wind - power and photovoltaic - power output, and construct a prediction data set for wind - power and photovoltaic - power output.

[0102] Step 2: Express the predicted quantity as the sum of a deterministic relationship and a noise term, as shown in the following formula:

[0103] y i =f(x i )+ε0(xi )

[0104] Among them, the deterministic relationship f(x i ) can be estimated through neural network training, as shown in the following formula:

[0105]

[0106] In the formula, y i represents the true value of the predicted quantity (the true value of the prediction target, and the prediction target includes new energy output and load), f(x i ) and respectively represent the deterministic relationship between the true value of the predicted quantity and its corresponding prediction factor and the deterministic relationship estimated through the neural network, ε0(x i ) and ε1(x i ) represent the noise of the sequence itself (sequence noise) and the error of the neural network estimation (prediction error), is the variance of the sequence noise, is the variance of the prediction error.

[0107] Step 3: Calculate and

[0108] based on the bootstrap theory and the neural network model, which specifically includes the following steps:

[0109] ① Select multiple neural network models as alternatives and label them according to Model 1, Model 2, etc., such as BP network, random forest (RF), support vector machine (SVM), extreme learning machine (ELM), LSTM, etc.;

[0110] ② Set the number of resampling times m, and perform m random sampling with replacement in the training dataset to obtain m bootstrap training sets. To ensure the prediction accuracy, the sampling quantity should be higher than 70% of the length of the training dataset;

[0111] ③ Input the m bootstrap training sets into Model 1 for training in sequence to obtain m sets of prediction results and m sets of errors;

[0112] ④ Based on the normal distribution assumption of the error and noise in the bootstrap theory, apply the histogram and Shapiro-Wilk test (Wtest) to perform a normal test on the prediction results and errors. If the test fails, replace the neural network model and repeat ③. After passing the test, go to ⑤;

[0113] ⑤Set the acceptable prediction accuracy threshold, and screen out the prediction results with a prediction accuracy higher than the threshold from the m groups of prediction results to form a set Ω. Based on the bootstrap theory, the average integration of the prediction results in Ω constitutes an unbiased estimate, as shown in the following formula:

[0114]

[0115] The uncertainty of the prediction model is represented by the variance of the prediction error, as shown in the following formula:

[0116]

[0117] In the formula, K is the number of elements in the set Ω, represents the prediction result of the k-th bootstrap training set;

[0118] ⑥Calculate the variance of the sequence noise that reflects the uncertainty of the prediction sequence itself through the true value, the integrated prediction result, and the model variance as shown in the following formula:

[0119]

[0120] And considering the non-negative property of the variance, process the negative values in the calculation results, as shown in the following formula:

[0121]

[0122] Step 4: Take the confidence level α and construct the prediction intervals for the wind power and photovoltaic power outputs.

[0123] Specifically, it includes:

[0124] ①Compare and Take the critical value of the standard normal distribution at the confidence level α. The precision level is 10 -q , and judge whether holds to determine whether holds.

[0125] Specifically, it is deduced from that when holds, also holds.

[0126] ②When holds, the noise error can be ignored, and the upper and lower bounds of the prediction interval are constructed as shown in the following formula. Otherwise, go to step ③;

[0127]

[0128]

[0129] ③ Re - establish the regression model using the sample data to establish the relationship with the input x i and predict and construct the upper and lower bounds of the prediction interval based on the prediction results as shown in the following formula:

[0130]

[0131] Step Five: Load interval prediction. Specifically, combine meteorology, holiday attributes, and seasonal characteristics to construct the load prediction dataset D, and construct the prediction interval of the load according to the methods in Steps Two to Four.

[0132] Step Six: Establish typical scenarios of the micro - grid. Based on the prediction intervals of wind power, photovoltaic power, and load, construct the maximum, minimum, and two strongest fluctuation scenarios of the total new - energy output and load respectively. Combine the new - energy scenarios and load scenarios pairwise to form 16 typical scenarios.

[0133] Step Seven: Take the minimum of the new - energy curtailment under 16 typical scenarios as the goal, and take the energy - storage charging power as the decision variable to establish an optimal dispatching model of the micro - grid. The objective function is the minimum of the new - energy curtailment under 16 typical scenarios, as shown in the following formula:

[0134]

[0135] Sort out the constraint conditions for the operation of the micro - grid as shown in the following formula:

[0136] Power - balance constraint:

[0137]

[0138] New - energy output constraint:

[0139]

[0140] Energy - storage state constraint:

[0141]

[0142] In the formula, is the new - energy curtailment power of the micro - grid; and respectively represent the new - energy output, load power, energy - storage charging power (the energy - storage charging power is negative for discharging and positive for charging), and grid - supply power of the micro - grid at time t under typical scenario s; respectively represent the predicted power generation of the total new - energy, wind power, and photovoltaic power at time t under typical scenario s; and respectively represent the maximum output powers of wind power and photovoltaic power in the microgrid; and respectively represent the allowable minimum and maximum electricity quantities of the microgrid energy storage and the charging electricity quantity at time t in the typical scenario s; Δt is the time interval from t - 1 to t; is the maximum value of the charge / discharge power of the energy storage in the microgrid, that is, under the charging condition, is the maximum value of the charging power of the energy storage in the microgrid, and under the discharging condition, is the maximum value of the discharging power of the energy storage in the microgrid.

[0143] Step Eight: Perform model solution through intelligent optimization algorithms such as particle swarm optimization to obtain the optimal operation process of the microgrid.

[0144] To address the impact of the uncertainty of new - energy output and load on the operation safety and benefits of the microgrid, the present invention provides a microgrid optimal scheduling method considering the dual uncertainties of new - energy output and load. This method quantifies the inherent uncertainty and prediction uncertainty of new - energy output and load based on the bootstrap theory and artificial neural network, constructs the prediction intervals of new - energy output and load, and obtains 16 typical source - load scenarios of the microgrid within the prediction intervals through pairwise combinations of the maximum, minimum, and two extreme fluctuation processes of new - energy and load. Taking the minimum of new - energy curtailment in the 16 extreme scenarios as the objective, an optimal scheduling model of the microgrid is established. The solution results can ensure the new - energy consumption level of the microgrid under uncertain scenarios. By quantifying the uncertainty of new - energy output and load in the microgrid through interval prediction, an optimal scheduling model considering uncertainty is further built and solved. The obtained operation process can ensure the new - energy consumption level and system operation efficiency under multiple uncertainty scenarios of new - energy output and load, thereby improving the operation stability of the microgrid.

[0145] An embodiment of a microgrid optimal scheduling system:

[0146] A microgrid optimal scheduling system includes a processor, and the processor is used to execute a computer program to implement the steps of a microgrid optimal scheduling method. Among them, a microgrid optimal scheduling method has been described in detail in an embodiment of a microgrid optimal scheduling method, and will not be elaborated here.

Claims

1. A microgrid optimal scheduling method, characterized in that, It includes the following steps: Establish a microgrid optimal scheduling model with the minimum new energy curtailment in a certain typical scenario within the prediction interval of the predicted quantity or the minimum sum of new energy curtailment in two or more typical scenarios as the objective, and the charge and discharge power of energy storage as the decision variable, and solve the model to obtain the microgrid optimal scheduling plan; When the condition of ignoring the sequence noise of the predicted quantity sequence holds, the prediction interval is taken as: Otherwise, it is taken as: Among them, and are respectively corresponding to the predicted quantity: the estimated deterministic relationship, the variance of the sequence noise, and the variance of the prediction error. The estimated deterministic relationship is the estimated deterministic relationship between the true value of the predicted quantity and the corresponding predictor. The true value of the predicted quantity is equal to the sum of the following for the predicted quantity: the estimated deterministic relationship, the sequence noise, and the prediction error. The sequence noise follows a normal distribution with a mean of 0 and a fixed variance of , and the prediction error follows a normal distribution with a mean of 0 and a fixed variance of . The predicted quantity is the uncertainty factor faced during the operation of the microgrid, is the critical value of the standard normal distribution at the confidence level α.

2. The microgrid optimal scheduling method according to claim 1, wherein The sequence noise ignoring condition is: Or where 10 -q is the precision level of the critical value.

3. The microgrid optimal scheduling method according to claim 1, wherein, and Calculated based on the predicted data set of predicted values, the bootstrap theory, and the neural network model.

4. The microgrid optimal scheduling method according to claim 1 or 3, characterized in that, The calculation formula is as follows: Among them, represents the prediction result of the k-th group of training sets. K is the number of elements in the set Ω. The prediction results with a prediction accuracy higher than the threshold are selected from the m groups of prediction results to form the set Ω.

5. The microgrid optimal scheduling method according to claim 1 or 3, characterized in that The calculation formula is as follows: where y i is the true value.

6. The microgrid optimal scheduling method according to claim 1 or 3, characterized in that The calculation formula is as follows: Among them, represents the prediction result of the k-th group of training sets. K is the number of elements in the set Ω. The prediction results with a prediction accuracy higher than the threshold are selected from the m groups of prediction results to form the set Ω.

7. The microgrid optimal scheduling method according to claim 1, characterized in that, The uncertainty factors include at least new energy output and load.

8. The microgrid optimal scheduling method according to claim 7, characterized in that The typical scenario is a combination of a new energy scenario and a load scenario. The new energy scenario includes the maximum, minimum, and strongest fluctuation of new energy output, and the load scenario includes the maximum, minimum, and strongest fluctuation of load.

9. The microgrid optimal scheduling method according to claim 8, wherein, The strongest fluctuation of new energy output is that the adjacent new energy output prediction points respectively take the minimum and maximum values within their corresponding prediction intervals; the strongest fluctuation of load is that the adjacent load prediction points respectively take the minimum and maximum values within their corresponding prediction intervals.

10. A microgrid optimal scheduling system, including a processor, characterized in that, The processor is used to execute a computer program to implement the steps of the microgrid optimal scheduling method according to any one of claims 1 to 9.

Images