A virtual power plant energy storage scheduling method and system and a storage medium

Abstract

The present application relates to the technical field of power system dispatching control, and discloses a virtual power plant energy storage dispatching method, system and storage medium, the method comprising: obtaining load historical prediction data and historical real load data of a virtual power plant; calculating load error based on the load historical prediction data and the historical real load data and performing distribution fitting; generating probability multi-scenario load data based on the fitting result; and using an optimization algorithm to optimize and solve the probability multi-scenario load data to obtain an energy storage dispatching scheme. The optimization algorithm uses a two-stage optimization algorithm, and the energy storage dispatching scheme is generated by drawing a box plot according to the dispatching time step of the optimization solution and performing statistical analysis. The present application improves the adaptability and stability of the energy storage dispatching scheme and reduces the operation cost by analyzing the historical prediction error to construct a probability multi-scenario and combining the two-stage optimization algorithm, thereby providing effective technical support for the long-term stable operation of the virtual power plant under the peak-valley time-of-use electricity price policy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system dispatching control, in particular to a virtual power plant energy storage dispatching method and system and a storage medium. BACKGROUND

[0002] As a new type of power resource management technology, virtual power plant integrates dispersed energy resources and loads to form a virtual power production and dispatching system. Energy storage system is an important part of virtual power plant, and the optimization of its dispatching strategy is of great significance to improve the economy and stability of the entire system.

[0003] The current virtual power plant energy storage dispatching method mainly has the following technical problems:

[0004] The existing energy storage dispatching method has deficiencies in dealing with load uncertainty. Traditional dispatching methods are usually optimized based on deterministic load prediction data, without fully considering the influence of load prediction error, resulting in poor adaptability of the dispatching scheme in actual operation. When there is a deviation between actual load and predicted load, the preset dispatching scheme often cannot effectively respond, affecting the operation efficiency of the virtual power plant.

[0005] The existing optimization algorithm has limitations in solving the energy storage dispatching problem. When using a single solver to solve, the solver only optimizes from a mathematical point of view, tending to fully charge the energy storage device at low price periods and fully discharge it at peak periods, resulting in frequent deep charging and discharging of the battery, frequent start and stop of the energy storage device, and single calculation results, lacking flexibility. When using an optimization algorithm alone, its calculation process usually has large fluctuations and is sensitive to load randomness, with low initial solution quality and swinging energy storage charging and discharging selection in some periods, affecting the stability of the dispatching scheme.

[0006] The existing method lacks in-depth analysis and use of historical prediction errors. Most dispatching methods ignore the regularity information contained in historical prediction errors and fail to effectively use these information to improve dispatching decisions, resulting in dispatching schemes that cannot fully reflect the probability distribution characteristics of actual load.

[0007] The long-term applicability of the existing dispatching scheme is insufficient. Traditional methods are usually optimized for specific periods or specific load scenarios, lacking comprehensive consideration of multiple possible scenarios, making the robustness of the dispatching scheme poor in long-term operation, requiring frequent adjustment and correction, increasing operation and maintenance costs.

[0008] These technical problems result in poor performance of the existing virtual power plant energy storage dispatching method in the face of load uncertainty, and the economy and stability of the dispatching scheme need to be improved, limiting the further promotion and application of virtual power plant technology. SUMMARY

[0009] The application aims to provide a virtual power plant energy storage scheduling method, system and storage medium, and aims to solve the problems of insufficient adaptability of the existing energy storage scheduling scheme and limited processing capacity for load uncertainty.

[0010] To solve the above technical problems, the application provides a virtual power plant energy storage scheduling method, comprising:

[0011] obtaining load historical prediction data and historical real load data of a virtual power plant;

[0012] calculating load error based on the load historical prediction data and the historical real load data and performing distribution fitting;

[0013] generating probability multi-scenario load data based on the fitting result; and

[0014] obtaining an energy storage scheduling scheme by using an optimization algorithm to optimize and solve the probability multi-scenario load data.

[0015] Optionally, the calculation formula of the load error is:

[0016] E load = L pred - L act ;

[0017] In the formula, E load is the load error, L pred is the load historical prediction data, and L act is the historical real load data.

[0018] The distribution fitting process comprises fitting an error distribution function to the load error.

[0019] The error distribution function comprises at least one of normal distribution, multivariate normal distribution or mixed normal distribution.

[0020] Optionally, the step of generating probability multi-scenario load data comprises:

[0021] sampling within a set confidence interval based on the result of the distribution fitting;

[0022] combining the sampling result with load typical values to form probability multi-scenario load data;

[0023] The sampling method comprises a Monte Carlo method or a Markov chain sampling method, and the load typical values are load historical average values or median values.

[0024] Optionally, the optimization algorithm is a two-stage optimization algorithm, comprising:

[0025] a first stage uses a mathematical programming solver to obtain an initial optimal solution;

[0026] The second stage uses a swarm intelligence optimization algorithm to obtain a global optimal solution based on the initial optimal solution.

[0027] Optionally, the step of obtaining the energy storage scheduling scheme comprises:

[0028] The optimization solution result is plotted as a box plot according to the scheduling time step;

[0029] A preliminary scheduling scheme is determined based on the statistical characteristics of the box plot; and

[0030] The preliminary scheduling scheme at time t in the scheduling period is corrected in combination with the constraint condition and the smaller value in the future predicted load to obtain a final energy storage scheduling scheme.

[0031] The statistical characteristics include at least one of a median, a maximum value, a minimum value and an outlier.

[0032] Optionally, the constraint condition in the optimization solution process comprises a power balance constraint, an energy storage charging and discharging power constraint, an energy storage state of charge constraint and a power transmission constraint of the power grid.

[0033] The objective function of the optimization solution is minimization of the virtual power plant operation cost.

[0034] The energy storage scheduling scheme is applicable to a peak-valley time-of-use electricity price policy.

[0035] Optionally, after the data is obtained, a data preprocessing step is further included:

[0036] Abnormal data is detected and processed.

[0037] Historical data is visually processed.

[0038] The abnormal data includes at least one of a missing value, a repeated value and an outlier.

[0039] Optionally, after the process of calculating the load error, the following is further included:

[0040] Using the preprocessed data, the time step of the load historical prediction data, the historical real load data and the load error is adjusted according to the scheduling step length requirement of the target virtual power plant.

[0041] The step length adjustment processing method includes one or more of a direct average method, an interpolation method, a data aggregation method or a time series prediction method.

[0042] The application further provides a virtual power plant energy storage scheduling system for implementing the above method, comprising:

[0043] A data acquisition module is configured to acquire historical load prediction data and historical real load data.

[0044] An error analysis module is configured to calculate load errors and perform distribution fitting.

[0045] A scenario generation module is configured to generate probabilistic multi-scenario load data based on the fitting results.

[0046] An optimization solution module is configured to solve by using an optimization algorithm.

[0047] A dispatch scheme generation module is configured to generate an energy storage dispatch scheme based on the optimization results.

[0048] The application further provides a computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the virtual power plant energy storage dispatch method.

[0049] Compared with the prior art, the application has at least the following beneficial effects:

[0050] The application can more accurately reflect the uncertainty characteristics of the load by establishing a probabilistic multi-scenario model based on historical errors, and avoids the limitations of traditional deterministic dispatch methods. The two-stage optimization algorithm combines the advantages of mathematical programming solvers and swarm intelligence optimization algorithms, which not only ensures the solution speed but also improves the solution quality, effectively solves the convergence and globality problems existing in single optimization methods, and significantly improves the adaptability and economy of the energy storage dispatch scheme.

[0051] The application extracts the common characteristics of the multi-scenario optimization results by the box plot statistical analysis method, generates an energy storage dispatch scheme with long-term applicability, and avoids the computational power expenditure of frequent recalculation. Combined with the constraint conditions of the peak-valley time-of-use electricity price policy, the dispatch scheme not only considers the load uncertainty but also fully utilizes the price difference, minimizes the operation cost under the premise of ensuring the safe and stable operation of the system, and provides effective technical support for the actual operation of the virtual power plant. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 A virtual power plant energy storage dispatch method step flowchart in an embodiment of the application;

[0053] Figure 2 A virtual power plant energy storage dispatch method step flowchart in another embodiment of the application;

[0054] Figure 3 A virtual power plant energy storage dispatch system module diagram in an embodiment of the application. DETAILED DESCRIPTION

[0055] The virtual power plant energy storage scheduling method, system and storage medium of the present application will be described in more detail below in conjunction with the schematic drawings, in which the preferred embodiments of the present application are represented, it should be understood that the present application described herein can be modified by those skilled in the art, while still achieving the advantageous effects of the present application. Therefore, the following description should be understood as a broad knowledge for those skilled in the art, and not as a limitation of the present application.

[0056] The present application will be described in more detail in the following paragraphs with reference to the accompanying drawings. The advantages and features of the present application will be more apparent from the following description and claims. It should be noted that the drawings are very simplified and all use non-precise proportions, only for the purpose of facilitating, clarifying the purpose of assisting the description of the embodiments of the present application.

[0057] The present application provides a virtual power plant energy storage scheduling method, please refer to Figure 1 , comprising the following steps:

[0058] S11, obtaining the load historical prediction data and historical real load data of the virtual power plant;

[0059] S12, calculating the load error based on the load historical prediction data and historical real load data and performing distribution fitting;

[0060] S13, generating probability multi-scenario load data based on the fitting result;

[0061] S14, using an optimization algorithm to optimize and solve the probability multi-scenario load data, and obtaining an energy storage scheduling scheme.

[0062] The virtual power plant energy storage scheduling method provided by the present embodiment, by analyzing and utilizing the load historical prediction error, not only improves the applicability of energy storage scheduling, but also helps to reduce operation and maintenance cost, improve decision efficiency and enhance data utilization efficiency.

[0063] In step S11, the load historical prediction data L pred and the historical real load data L act of the target virtual power plant in the historical time period are selected.

[0064] The load historical prediction data can be derived from one or a combination of machine learning models, probability models, and physical models. The historical time period should be long enough to ensure the representativeness of the data, usually 1-3 years of historical data are selected.

[0065] In one specific example, the load historical prediction data and the corresponding actual load data at every 15-minute interval in the past two years can be selected to form a complete historical data set.

[0066] In step S12, the load error is first calculated, and the calculation formula is:

[0067] E load = L pred - L act ;

[0068] In the formula, E load For load error, L pred For historical load forecast data, L act This is historical, actual load data.

[0069] The process of fitting the distribution includes: fitting the load error with an error distribution function.

[0070] Specifically, the error distribution function includes at least one of a normal distribution, a multivariate normal distribution, or a mixed normal distribution.

[0071] Taking the normal distribution as an example, its expression is:

[0072] ;

[0073] in, Let be the mean and standard deviation of the error normal distribution, respectively, and let the distribution function be:

[0074] ;

[0075] in, This represents the active power value for the load error.

[0076] In this embodiment, through in-depth analysis of historical prediction errors, the relationship between the virtual power plant prediction model and the actual load can be captured more accurately, laying the foundation for subsequent uncertainty modeling.

[0077] In step S13, the error is sampled to obtain multiple sets of values ​​with the same distribution as the load error, which are then combined with typical load values ​​within the scheduling period to obtain probabilistic multi-scenario results for subsequent optimization.

[0078] The step of generating probabilistic multi-scenario load data based on the fitting results specifically includes:

[0079] S131. Based on the distribution fitting results, sampling is performed within a set confidence interval;

[0080] Using the error distribution function f(E) obtained after fitting load Select the confidence interval [E] lower E upper The load error compensation value is obtained by performing multiple samplings within the confidence interval. The superscript i represents the group.

[0081] In one specific example, the sampling method includes a Monte Carlo method or a Markov chain sampling. The Monte Carlo method is preferred because of its simplicity and good convergence.

[0082] In step S132, the sampling result is combined with the load typical value to form the probabilistic multi-scenario load data.

[0083] The sampled value is combined with the load typical value L typical to form the probabilistic multi-scenario load matrix containing uncertainty :

[0084] ;

[0085] In one specific example, the load typical value is a load history average value or a median value. The load history average value can reflect the long-term load level, and the median value has better robustness to abnormal values.

[0086] In this embodiment, the sampling is performed by introducing historical prediction errors, which actually simulates various deviation conditions that may occur in reality, so that the model can better cope with future uncertainties, rather than only for a specific date.

[0087] Further, the probabilistic multi-scenario load data is optimized and solved by using an optimization algorithm to obtain a storage scheduling scheme.

[0088] In step S14, the optimization algorithm is a two-stage optimization algorithm, which includes:

[0089] The first stage uses a mathematical programming solver to obtain an initial optimal solution;

[0090] The is solved as the initial input of the optimization algorithm, specifically: the probabilistic multi-scenario load matrix is grouped and solved by a mathematical programming solver to obtain the optimal initial scheduling solution of each scenario under the constraint of the target function .

[0091] In one specific example, the mathematical programming solver can be a Cplex, Gurobi, etc. solver, or an open-source linear programming or mixed integer programming solver.

[0092] In one specific example, the Cplex solver is used to solve the initial optimal solution matrix of the target scheduling under the constraint condition . Cplex is a commercial optimization engine developed by IBM, which is commonly used to solve dynamic scheduling problems in optimal control, and is used in virtual power plant storage scheduling to quickly calculate the initial solution of storage charging and discharging.

[0093] Further, the second stage uses a swarm intelligence optimization algorithm to obtain a global optimal solution based on the initial optimal solution.

[0094] An initial scheduling solution is obtained An initialization condition is set according to the swarm intelligence optimization algorithm to select a partial solution space, and the selected solution space is used as an initial input for the position of part of individuals in the swarm intelligence optimization algorithm, and the remaining positions are randomly generated. Bayesian optimization is used for global optimization to obtain an optimal scheduling solution of the optimal objective function under constraints.

[0095] In a specific example, the swarm intelligence optimization algorithm can be a whale optimization algorithm (WOA), a genetic algorithm, a particle swarm algorithm, etc.

[0096] Taking the whale optimization algorithm as an example, the algorithm dimension is the same as the total number of time steps in the period to be solved, and the last dimension satisfies the virtual power plant constraint algorithm, and the other dimensions are independent of each other. Each dimension parameter adjustment can use Bayesian optimization for global optimization to obtain an optimal scheduling solution of the optimal objective function under constraints Each dimension represents a time interval corresponding to its order in the energy storage scheduling scheme.

[0097] After the optimization process of each dimension of the original swarm intelligence optimization algorithm, one or more of the following methods can be appropriately added according to the convergence speed and result satisfaction: crossover mutation and simulated annealing. It should be noted that crossover mutation and simulated annealing can help improve search efficiency and solution quality. Crossover mutation is mainly used to increase the diversity of solutions and promote the combination of excellent characteristics, and is suitable for complex problems that require large-scale parallel search; simulated annealing provides an effective mechanism to avoid falling into a local optimal solution, and is suitable for problems with very large solution spaces and many local extreme points. Therefore, in a specific example, one or more of the above methods can be added during WOA update iteration.

[0098] In this embodiment, the core technical effect of the two-stage optimization algorithm is that it combines the fast calculation speed of mathematical programming solvers and the excellent global performance of swarm intelligence optimization algorithms, ensuring both solution efficiency and solution quality, and avoiding the limitations of a single method.

[0099] Further, the calculation process needs to consider constraints, including system constraints in the virtual power plant, thermal power plant constraints, and energy storage constraints.

[0100] In the optimization solving process in this embodiment, constraints are set.

[0101] The constraint conditions in the optimization solution process include: power balance constraint, energy storage charging and discharging power constraint, energy storage state of charge constraint and power grid power transmission constraint.

[0102] Power balance constraint: power balance is the basic condition to ensure the operation of virtual power plant:

[0103] ;

[0104] Where, P pv,t is the power generation of photovoltaic power station at time t; P load,t is the load value at time t; P bat,t and |P batc,t | are the charging and discharging power of energy storage at time t (defining the discharging of energy storage as positive value and the charging as negative value); P mt-n,t is the output power of the nth thermal power unit at time t; N mt is the number of thermal power units; is a Boolean variable, which represents discharging when it is 1 and charging when it is 0.

[0105] Energy storage constraint: energy storage constraint is mainly limited by itself and operation requirements, and its main purpose is to prolong the service life:

[0106] SOC min ≤ SOC t ≤ SOC max ;

[0107] 0 ≤ P batd,t ≤ P batd max,t ;

[0108] 0 ≤|P batc,t | ≤ |P batc max,t |;

[0109] E storage (0) = E storage (T);

[0110] P batd,t ≤ P load,t ;

[0111] Where, SOC t represents the current state of charge of energy storage; SOC min and SOC max are the minimum and maximum state of charge set for protecting energy storage; P batd max,t and P batc max,t respectively represent the maximum charging and discharging power of energy storage system; E storage (0) and E storage (T) represent the initial state and the state at the end of the dispatching period of energy storage system.

[0112] (3) Power grid power transmission constraints:

[0113] P bat,t ≤ P line,max ; |P batc,t| ≤ |P line,max |

[0114] where P line,max is the maximum value of power exchange between the virtual power plant and the distribution network tie line.

[0115] (4) Thermal power plant constraints mainly include output constraints, ramp constraints, and reserve capacity constraints.

[0116] Specifically, the output constraints are:

[0117] ;

[0118] where P mt_min is the minimum value of the power output of the nth thermal power unit, P mt_max is the maximum value of the power output of the nth thermal power unit, U mt-n_t represents the operating state of the nth thermal power unit at time t, and is 1 if it is running and 0 if it is stopped.

[0119] Specifically, the reserve capacity constraints of the thermal power unit are:

[0120] , ;

[0121] where and are the lower reserve capacity and upper reserve capacity of the nth unit at time t.

[0122] Specifically, the reserve capacity of the thermal power unit is subject to the ramp limit constraints:

[0123] ;

[0124] ;

[0125] where r cp,down and r cp,up are the upper and lower ramp rates, is the response time of the reserve, usually 10 minutes.

[0126] Specifically, the reserve capacity of the thermal power unit is subject to the maximum output limit constraints: , ;

[0127] where the parameters have been explained above.

[0128] Further, the objective function is set:

[0129] The objective function of the optimization solution is the minimization of virtual power plant operation cost:

[0130] ;

[0131] Wherein, C pv is the photovoltaic power station power generation price, C grid,t is the time-of-use electricity price, C bat is the storage charging and discharging operation cost unit price, b n is the power generation coefficient, S n is the unit start-stop coefficient.

[0132] It should be noted that the above formula includes seven items of addition, wherein the first item is the photovoltaic power station power generation cost, the second item is the storage grid charging cost, the third and fourth items are the storage charging and discharging cost, the fifth item is the grid direct power supply cost to the load, and the sixth and seventh items are the fuel cost and start-stop cost of the thermal power unit.

[0133] The storage scheduling scheme is applicable to the peak-valley time-of-use electricity price policy. Under the peak-valley time-of-use electricity price policy, the storage system charges in the low price valley period and discharges in the high price peak period, so as to realize the peak load shifting and economic benefit maximization of the power system.

[0134] In step S14, the step of obtaining the storage scheduling scheme specifically includes:

[0135] S141, statistical analysis of the optimization solution result;

[0136] S142, generating a storage scheduling scheme based on the statistical analysis result.

[0137] The statistical analysis in the process of S141 specifically includes:

[0138] S1411, drawing a box plot according to the scheduling time step length based on the optimization solution result;

[0139] S1412, determining a preliminary scheduling scheme based on the statistical characteristics of the box plot;

[0140] S1413, modifying the preliminary scheduling scheme in combination with the constraint condition to obtain the final storage scheduling scheme.

[0141] Specifically, the optimal solution obtained by the group intelligent optimization algorithm is drawn into a box plot according to the peak-valley time-of-use electricity price time interval and the scheduling time step length, and a preliminary storage scheduling scheme P opt for the virtual power plant is obtained in combination with the statistical characteristics shown by the box plot.

[0142] In step S1412, the statistical characteristics include at least one of a median, a maximum value, a minimum value, and a discrete point. In one specific example, the median of the box plot can be taken as the current step target scheduling value, or the average value of the total amount of the electricity price time period can be calculated as the scheduling value of each time step.

[0143] In the preliminary scheduling scheme, the energy storage charging power has a positive correlation in the same electricity price time period, and the energy storage discharging power should be selected in the scheduling period t under the condition of meeting the constraints and combining the subsequent predicted load data. opt,t and the future predicted load L future,t The smaller value is:

[0144] The adjustment is performed, so as to obtain the energy storage scheduling scheme applicable to the long-term operation of the target virtual power plant under the peak-valley time-of-use electricity price and the long-term stable operation of the load characteristics. .

[0145] Among them, the average value method or the median can be used for the selective adjustment of the electricity price continuous time period; and the subsequent predicted load can be provided by a higher-precision real-time intra-day prediction.

[0146] It should be noted that the scheme is a day-ahead scheduling scheme of the energy storage under the peak-valley time-of-use electricity price, and the power grid only faces the virtual power plant for power supply and does not accept power transmission. The future predicted load power can be updated and adjusted in real time by the day-ahead load prediction power and the intra-day load prediction power, and a more accurate intra-day prediction can better support the scheduling scheme.

[0147] When the load characteristics are one or a mixture of Gaussian distribution, logarithmic distribution, and gamma distribution, and the future distribution will not change significantly due to external factors (such as technological changes, climate changes, and policy interventions), the method is applicable.

[0148] In the embodiment, the core technical effect of the step is that the energy storage scheduling of the target virtual power plant under the peak-valley time-of-use electricity price obtained by the advantages of the box plot has long-term applicability, which is beneficial to the long-term operation of the virtual power plant, and the segmented scheduling scheme can be controlled to provide help for the intra-day real-time adjustment scheduling with higher precision requirements.

[0149] In the embodiment of the present application, by using the scheme of generating a probability multi-scenario, the virtual power plant is closer to the daily load demand, the Cplex-WOA two-stage optimization method is built, the Cplex has the characteristics of fast calculation speed and the WOA has the characteristics of excellent global performance, the energy storage scheduling of the target virtual power plant under the peak-valley time-of-use electricity price obtained by the advantages of the box plot has long-term applicability, which is beneficial to the long-term operation of the virtual power plant, and the segmented scheduling scheme can be controlled to provide help for the intra-day real-time adjustment scheduling with higher precision requirements.

[0150] Further, in another specific example, after the data is acquired in step S11, a data preprocessing step is further included:

[0151] The data preprocessing process includes:

[0152] detecting and processing abnormal data;

[0153] visualizing the historical data.

[0154] The abnormal data includes at least one of missing values, duplicate values, and outliers.

[0155] The visualization includes one or more of scatter plots, histograms, trend charts, or probability density distribution charts.

[0156] The specific methods of data cleaning include deleting duplicate data, box plot detecting abnormal data, random forest regression filling missing data, or date format conversion, which are common means for those skilled in the art and will not be described in detail.

[0157] Further, in one specific example, after the process of calculating the load error in step S12, a time step adjustment is further included.

[0158] Using the preprocessed data, the time steps of the load history prediction data, the historical real load data, and the load error are adjusted according to the scheduling step requirement of the target virtual power plant.

[0159] The adjusted load error is then fitted to obtain the distribution of the load error.

[0160] The step adjustment method includes one or more of direct averaging, interpolation, data aggregation, or time series prediction.

[0161] For example, to adjust 15-minute interval data to 1-hour interval scheduling steps, the direct averaging method can be used to take the average of 4 15-minute data, or the interpolation method can be used for time alignment.

[0162] In summary, the embodiments of this invention utilize historical load forecasting error analysis to enable the model to better capture the relationship between the virtual power plant forecasting model and the actual load. By generating numerous scenarios to simulate modified real loads with uncertainties, the maximum probable solution close to the actual load condition is obtained. Through simulations of numerous scenarios, the operating patterns of the virtual power plant under peak-valley time-of-use pricing are obtained, and box plot analysis reveals the energy storage dispatch patterns for a single time step and a specific price range. This method helps the model exhibit better robustness when facing noisy data or outliers. Energy storage dispatch is crucial for power system scheduling and operation. This solution summarizes the energy storage dispatch patterns corresponding to the target virtual power plant through scenario generation, further helping the virtual power plant to more effectively schedule and manage resources, reducing real-time scheme corrections caused by load fluctuations and uncertainties. Compared to intraday real-time adjustments, only specific price time intervals need to be modified, saving computational overhead. Compared to single-day dispatch, it improves the long-term applicability of dispatch, significantly reducing the uncertainty caused by load fluctuations, thereby reducing grid operation and maintenance costs and improving decision-making efficiency.

[0163] Example 2

[0164] Please refer to Figure 2 This embodiment provides a specific implementation process for a virtual power plant energy storage dispatching method based on Embodiment 1:

[0165] S21: Select historical load power data and corresponding historical predicted load power data within a certain time period.

[0166] For example, select historical load forecast data and corresponding historical load forecast power data for a virtual power plant over a two-year period from January 2022 to January 2024. This virtual power plant includes a 50MW photovoltaic power station, a 20MW energy storage system, and several thermal power units. Data collection intervals are 15 minutes. The acquired historical load forecast data L... pred Derived from a load forecasting model based on an LSTM neural network, historical real load data L act Actual measurement values ​​derived from the SCADA system.

[0167] S22: Perform data visualization operations, check for missing values, duplicate values, or other abnormal data, and clean the data.

[0168] For example, the acquired two years of historical data were first visualized to create scatter plots, trend charts, and probability density distribution maps. The analysis revealed approximately 0.3% missing values, 0.1% duplicate values, and some outliers due to sensor malfunctions.

[0169] Data cleaning steps include:

[0170] Outliers are identified using the boxplot detection method, and data exceeding 1.5 times the interquartile range is marked as an outlier.

[0171] Duplicate data records are deleted.

[0172] Random forest regression method is used to fill in missing data, and load data at previous and subsequent times and auxiliary variables such as weather and temperature are used for prediction and filling.

[0173] All timestamps are converted to a standard date format.

[0174] S23: Calculate the difference between historical data and historical prediction data, adjust the time step, and fit the error distribution.

[0175] S24: Use the distribution formed by combining error distribution and uncertainty to select a confidence interval for random sampling to obtain secondary prediction sample data.

[0176] S25: Call Cplex to solve the initial solution of each group of samples, and input the multi-scenario load data into the Cplex solver.

[0177] An optimization model is established to minimize the operating cost of the virtual power plant:

[0178] The objective function is:

[0179]

[0180] The constraints include power balance constraints, energy storage SOC constraints (20% to 90%), and charge and discharge power constraints (±10 MW).

[0181] S26: Set the WOA parameter range, and select the initial solution according to the conditions for initialization and Bayesian parameter optimization.

[0182] S27: Perform fitness calculation and dynamic mutation to obtain the optimal solution under the global optimal parameters.

[0183] The WOA algorithm calculates the fitness value in each iteration, and uses a dynamic mutation strategy to improve the global search ability of the algorithm. When there is no improvement for consecutive generations, a simulated annealing mechanism is introduced to avoid premature convergence.

[0184] S28: The optimal solution under the global optimal parameters can be visualized, and the boxplot is drawn according to the time step.

[0185] Each hour corresponds to a boxplot, which displays the median, quartile, maximum, minimum, and scatter point of the energy storage dispatch power at that time.

[0186] For example, analyzing the boxplot found that:

[0187] 02:00-06:00 (low tariff period): the energy storage mainly charges, with a median of -6.5 MW;

[0188] 08:00-11:00 and 18:00-21:00 (peak tariff period): the energy storage mainly discharges, with a median of +8.2 MW;

[0189] Other periods (flat tariff period): the energy storage is in standby state, with a median close to 0 MW.

[0190] S29: According to the visualized box plot, the time intervals are divided into corresponding charging and discharging time intervals according to the peak-valley time-of-use electricity price.

[0191] For example, according to the local peak-valley time-of-use electricity price policy and the analysis results of the box plot, 24 hours are divided into three time intervals:

[0192] Charging time interval: 22:00-06:00 the next day (low tariff 0.35 yuan / kWh);

[0193] Discharging time interval: 08:00-11:00, 18:00-23:00 (peak tariff 0.85 yuan / kWh);

[0194] Standby time interval: 06:00-08:00, 11:00-18:00 (flat tariff 0.60 yuan / kWh).

[0195] S210: According to the selected method, the charging and discharging power at each time in each time interval is planned, and the discharging time interval is adjusted to a new energy storage planning scheme according to the judgment condition.

[0196] Based on the statistical characteristics of the box plot, a preliminary scheduling scheme is formulated:

[0197] Charging time interval: the median of the box plot at each time is taken as the charging power;

[0198] Discharging time interval: the median of the box plot at each time is taken as the discharging power reference value.

[0199] Considering the actual operation constraints, the scheme is modified:

[0200] The discharging power of the energy storage cannot exceed the real-time load demand, ensuring that the SOC of the energy storage returns to the initial state at the end of the scheduling period, and meeting the continuous charging and discharging time limit of the energy storage. Finally, a suitable energy storage scheduling scheme for long-term operation of the virtual power plant is obtained. Under the premise of ensuring the safe and stable operation of the system, the daily operation cost is reduced compared with the traditional scheduling method, and the utilization rate of the energy storage is improved.

[0201] Example Three

[0202] Please refer to Figure 3The embodiment of the present application provides a virtual power plant energy storage scheduling system for realizing the above method, comprising a data acquisition module, an error analysis module, a scenario generation module, an optimization solving module and a scheduling scheme generation module.

[0203] The data acquisition module is used for acquiring load historical prediction data and historical real load data. The module is connected with a data acquisition system of the virtual power plant, and extracts the load historical prediction data and the historical real load data from a historical database.

[0204] The error analysis module is used for calculating load errors and performing distribution fitting. The module is connected with the data acquisition module, receives the load data, calculates errors and performs statistical analysis and distribution fitting.

[0205] The scenario generation module is used for generating probability multi-scenario load data based on fitting results. The module is connected with the error analysis module, and generates multiple sets of scenario data according to the fitted error distribution function.

[0206] The optimization solving module is used for solving by using an optimization algorithm. The module comprises a mathematical programming solving submodule and a swarm intelligent optimization submodule, and the two submodules work cooperatively to realize two-stage optimization.

[0207] The scheduling scheme generation module is used for generating an energy storage scheduling scheme based on optimization results. The module is connected with the optimization solving module, performs statistical analysis on the optimization results and generates a final scheduling scheme.

[0208] The modules are connected through data interfaces to transmit information, and the system as a whole adopts a modular design, facilitating maintenance and upgrading.

[0209] The embodiment also provides a computer readable storage medium, and computer executable instructions are executed by a processor to realize the method embodiments of the present application. The computer readable storage medium includes permanent and non-permanent, removable and non-removable media, and can be realized by any method or technology. Information can be computer readable instructions, data structures, program modules or other data. Examples of the storage medium of the computer include but are not limited to phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tape, magnetic tape disk storage or other magnetic storage devices or any other non-transmission medium, which can be used to store information that can be accessed by a computing device.

[0210] The application makes the model better capture the relationship between the virtual power plant prediction model and the real load by using the load history prediction error analysis, and the model shows better robustness when facing noise data or outliers. The method of scene generation simulates a large number of corrected real loads containing uncertainties, thereby obtaining the maximum possible solution close to the real load condition, which helps the virtual power plant to more effectively schedule and manage resources and reduce the operation and maintenance cost of the power grid. Energy storage scheduling is crucial for power system scheduling and operation. The scheme summarizes the energy storage scheduling rule of the target virtual power plant through scene generation, reduces the real-time scheme correction caused by load fluctuation and uncertainty, and only needs to modify the specific time interval of the electricity price compared with the daily real-time adjustment, saving the computing power overhead. Compared with single day-ahead scheduling, it improves the long-term applicability of scheduling, greatly reduces the uncertainty caused by load fluctuation, and helps to improve decision efficiency and enhance data utilization efficiency. Combined with the characteristics of fast calculation speed of mathematical programming solver and excellent global performance of swarm intelligence optimization algorithm, the energy storage scheduling of the target virtual power plant under the peak-valley time-of-use electricity price obtained by the box plot has long-term applicability.

[0211] In summary, the application has important practical significance for promoting the development of virtual power plant energy storage and electricity market.

[0212] Obviously, those skilled in the art can make various modifications and variations to the application without departing from the spirit and scope of the application. Thus, if these modifications and variations of the application fall within the scope of the claims of the application and their equivalents, the application also intends to include these modifications and variations.

Claims

1. A virtual power plant energy storage dispatching method, characterized in that, The method comprises the following steps: obtaining load historical prediction data and historical real load data of a virtual power plant; calculating load errors based on the load historical prediction data and the historical real load data and performing distribution fitting; generating probability multi-scenario load data based on the fitting result; and using an optimization algorithm to optimize and solve the minimum virtual power plant operation cost and its constraint conditions based on the probability multi-scenario load data, listing scenarios in a matrix, calculating each scenario in the matrix, and obtaining an energy storage scheduling scheme; wherein the optimization algorithm is a two-stage optimization algorithm, comprising: a first stage using a mathematical programming solver to obtain an initial optimal solution; a second stage using a swarm intelligence optimization algorithm to obtain a global optimal solution based on the initial optimal solution; the first stage and the second stage include an optimization objective, and share the same objective function and constraint conditions; the step of obtaining the energy storage scheduling scheme comprises: drawing a box plot according to the optimization solution, the peak-valley time-of-use electricity price time interval, and the scheduling time step; determining a preliminary scheduling scheme based on the statistical characteristics of the box plot; and combining the constraint conditions, selecting the preliminary scheduling scheme at time t in the scheduling period and the smaller value in the future predicted load to correct, and obtaining a final energy storage scheduling scheme; wherein the statistical characteristics include at least one of the median, the maximum value, the minimum value, and the discrete point.

2. The virtual power plant energy storage dispatching method according to claim 1, characterized in that, The calculation formula of the load error is: E load = L pred - L act ; In the formula, E load is the load error, L pred is the load history prediction data, L act is the historical real load data; The distribution fitting process comprises fitting the load error with an error distribution function; wherein the error distribution function comprises at least one of a normal distribution, a multivariate normal distribution, or a mixed normal distribution.

3. The virtual power plant energy storage dispatching method according to claim 1, characterized in that, The step of generating probability multi-scenario load data comprises: sampling within a set confidence interval based on the result of the distribution fitting; combining the sampling result with load typical values to form probability multi-scenario load data; wherein the sampling method comprises a Monte Carlo method or a Markov chain sampling method; and the load typical value is a load historical average value or a median.

4. The virtual power plant energy storage dispatching method according to claim 1, characterized in that, The constraint conditions in the optimization solving process include power balance constraints, energy storage charging and discharging power constraints, energy storage state of charge constraints, and power transmission constraints of the power grid; the objective function of the optimization solving is the minimization of the virtual power plant operation cost; the energy storage scheduling scheme is applicable to the peak-valley time-of-use electricity price policy.

5. The virtual power plant energy storage dispatching method according to claim 1, characterized in that, After obtaining the data, a data preprocessing step is further included: detecting and processing abnormal data; visualizing the historical data; wherein the abnormal data includes at least one of missing values, duplicate values, and outliers.

6. The virtual power plant energy storage scheduling method according to claim 5, wherein after the process of calculating the load error, the following steps are further included: using the preprocessed data, adjusting the time steps of the load historical prediction data, the historical real load data, and the load error according to the scheduling step requirement of the target virtual power plant; wherein the step adjustment processing method comprises one or more of a direct average method, an interpolation method, a data aggregation method, or a time series prediction method.

7. A virtual power plant energy storage dispatch system implementing the method of any one of claims 1 to 6, characterized by The method comprises the following steps: a data acquisition module for obtaining load historical prediction data and historical real load data; an error analysis module for calculating load errors and performing distribution fitting; a scenario generation module configured to generate probabilistic multi-scenario load data based on the fitting result; an optimization solution module configured to perform optimization solution on minimization of virtual power plant operation cost and constraint conditions thereof by using an optimization algorithm based on the probabilistic multi-scenario load data; a dispatch scheme generation module configured to determine a preliminary dispatch scheme based on box plot statistical features of the optimization solution result, and correct the preliminary dispatch scheme to obtain a final energy storage dispatch scheme in combination with constraint conditions and future predicted load.

8. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by a processor to implement the method of any one of claims 1 to 6.

Images