Probability Revenue Analysis Method and Readable Storage Medium Applicable to User-Side Microgrid

Abstract

A probabilistic benefit analysis method and a readable storage medium applicable to a user-side microgrid. The analysis method includes step S1: obtaining daily historical data and prediction data, clustering historical curves in the daily historical data to obtain daily types of load and renewable energy power generation, and selecting combinations of daily types of load and renewable energy power generation to form multiple daily typical types; step S2: selecting a daily typical type, and obtaining a set of simulation scenarios of the selected daily typical type according to the daily historical data and prediction data; step S3: repeating step S2 until sets of simulation scenarios of all daily typical types are obtained, and performing step S4: randomly sampling and calculating the benefits and the probabilities of the benefits according to the proportions of the respective daily typical types in the sets of simulation scenarios of the daily typical types. The present invention analyzes the uncertainty of the benefits of the user-side microgrid by transforming it into a probability problem.

Description

Technical Field

[0001] The present invention relates to the technology of microgrid revenue analysis, and particularly to a probabilistic revenue analysis method applicable to the user-side microgrid. Background Art

[0002] With the increasing depletion of fossil fuels, renewable energy has been promoted and applied in various countries around the world, and more and more distributed renewable energy sources are connected to the power system. Although distributed generation technology has advantages in energy conservation, emission reduction, and prevention of large-scale power outages, the large-scale access of distributed energy has caused problems such as increased difficulty in power grid operation and dispatch, deteriorated power quality, and impact on relay protection.

[0003] To coordinate the access of distributed energy, the microgrid, as a small power generation and distribution system capable of realizing self-control, protection, and management, has developed rapidly due to its advantages in flexibly absorbing renewable energy, increasing power supply reliability, and bringing economic benefits. For some large industrial and commercial users, it is necessary to evaluate the revenue before constructing a microgrid or putting into operation or expanding the equipment in the microgrid, so as to make risk decisions. However, compared with the large power grid, the microgrid has stronger uncertainty, specifically manifested as stronger load uncertainty, higher penetration rate of non-dispatchable resources such as solar energy and wind energy, and larger prediction errors of load / photovoltaic output / wind power output. Therefore, the revenue of the microgrid operation in the future for a period of time is uncertain, and the cost and revenue of the microgrid are affected by many factors, such as microgrid configuration parameters, load / photovoltaic output / wind power output conditions, electricity price levels, dispatch algorithms, etc. It is necessary to consider various factors in an overall manner and analyze and evaluate the revenue of the user-side microgrid through a reasonable and accurate method.

[0004] Currently, existing solutions all perform production simulations on the microgrid based on historical data to determine the revenue of the microgrid operation, which essentially belongs to deterministic analysis and has limited reference significance; at the same time, existing microgrid revenue analysis methods usually only consider the electricity cost in terms of electricity consumption cost, and do not consider the impact of demand charges or capacity charges in the actual two-part electricity price on the revenue. Summary of the Invention

[0005] The purpose of the present invention is to overcome the defects of the prior art and provide a reasonable and accurate probabilistic revenue analysis method applicable to the user-side microgrid.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions:

[0007] A probabilistic revenue analysis method applicable to the user-side microgrid includes the following steps:

[0008] S1: Obtain the historical data and predicted data for each day, cluster the historical curves in the daily historical data to obtain the daily types of load and renewable energy generation power, and select combinations of the daily types of load and renewable energy generation power to form multiple daily typical types;

[0009] S2: Select one daily typical type, and obtain the simulation scenario set of the selected daily typical type according to the historical data and predicted data for each day;

[0010] S3: Repeat step S2 until the simulation scenario sets of all daily typical types are obtained, and proceed to step S4;

[0011] S4: According to the proportions of each daily typical type, randomly sample and calculate the benefits and the probabilities of the benefits in the simulation scenario sets of each daily typical type.

[0012] Preferably, in step S1, the historical data includes the daily historical load curve and the historical curves of at least one renewable energy generation power, the predicted data includes the load prediction curve and the predicted curve of renewable energy generation power, cluster the historical load curve and the historical curves of renewable energy generation power respectively to obtain multiple daily types and the proportions of the daily types, select one daily type of load and its proportion and at least one daily type of renewable energy generation power and its proportion for free combination to obtain multiple daily typical types and their proportions.

[0013] Preferably, step S2 includes the following steps,

[0014] S2.1: Select one from multiple daily typical types, calculate the absolute prediction error according to the historical load curve, load prediction curve, historical curve of renewable energy generation power and its prediction curve related to the daily typical type, and obtain the prediction error probability model;

[0015] S2.2: In the selected daily typical type, randomly sample the historical load curve and the historical curves of renewable energy generation power respectively, and perform day-ahead scheduling using the prediction curves corresponding to the sampled samples;

[0016] S2.3: According to the prediction curve and the prediction error probability model, obtain the simulation scenario curve for real-time scheduling and perform a refined production simulation once;

[0017] S2.4: Repeat steps S2.2 - S2.3 N times until the simulation scenario set of the selected daily typical type is obtained.

[0018] Preferably, in step S1, the renewable energy power generation includes photovoltaic output and wind power output. If the daily load, photovoltaic output, and wind power output are independent of each other, and the proportions of each type of load, photovoltaic output, and wind power output remain unchanged in a future period of time, the proportions of each daily typical type are as follows:

[0019] P(l) = P load (i)·P pv (j)·P wind (k)

[0020] l = 1...n, i = 1...n load , j = 1...n pv , k = 1...n wind

[0021] where n is the number of daily typical types, n = n load ·n pv ·n wind ; n load , n pv , n wind are the numbers of daily historical curve types of load, photovoltaic output, and wind power output respectively.

[0022] Preferably, in step S2, the non-parametric kernel density estimation method is used to describe the probability distribution of the prediction error at each moment and the joint probability distribution of the prediction errors between adjacent moments.

[0023] Preferably, in step S2.3, an actual scheduling strategy is used for refined production simulation. The scheduling strategy includes two stages: day-ahead planning and real-time scheduling;

[0024] In the day-ahead planning stage, the whole day is divided into several time periods. Random sampling is carried out in the load prediction curve and the prediction curve of renewable energy power generation obtained in step S1. With the minimum operation cost of the microgrid throughout the day as the optimization goal, the day-ahead planning is modeled as a mixed-integer linear programming problem;

[0025] In the real-time scheduling stage, for the next optimization time period, using the current load and renewable energy power generation, with the minimum deviation between the microgrid devices and the day-ahead plan as the optimization goal, the real-time scheduling is modeled as a quadratic programming problem.

[0026] Preferably, in the real-time scheduling stage, the prediction error at each moment is simulated point by point. The load simulation value at each moment and the simulation value of the renewable energy power generation are obtained by subtracting the prediction error at each point from the load prediction curve and the renewable energy power generation prediction curve. Among them, the prediction error at the first moment is randomly sampled from the corresponding prediction error probability distribution, and the prediction errors at other moments are randomly sampled from the corresponding joint probability distribution of prediction errors after determining the prediction error at the previous moment.

[0027] Preferably, in step S4, according to the proportion of each daily typical type, the daily typical type of a certain day is randomly determined, and the load and renewable energy power generation prediction curves under this daily typical type are randomly selected from the simulation scenario set of this daily typical type for the day-ahead plan. Using the obtained error prediction distribution model, the prediction error at each moment is randomly sampled one by one to determine the prediction error at each moment, and the simulated load and renewable energy power generation curves are obtained based on the prediction curves for implementation scheduling; thus, the operation situation of the microgrid on a certain day is obtained, and the microgrid revenue is calculated in combination with the daily microgrid operation cost.

[0028] Preferably, in step S4, to calculate the revenue, the microgrid configuration parameters, operation mode, electricity price calculation method, and the historical curves and prediction curves of the load and renewable energy power generation of the microgrid for a certain period of time need to be given.

[0029] Preferably, in step S4, to calculate the revenue, the cost of the user-side microgrid needs to be considered. The cost of the user-side microgrid includes the converted equipment investment and operation and maintenance costs; the revenue of the user-side microgrid includes the revenue saved from the two-part electricity price and the revenue from selling surplus electricity to the grid.

[0030] The present invention discloses a probabilistic revenue analysis method applicable to the user-side microgrid, which analyzes the uncertainty of the revenue of the user-side microgrid by transforming it into a probability problem. First, the distribution of uncertain resources (load, photovoltaic power generation, wind power generation, etc.) in the microgrid on a long time scale is analyzed, and the influence of prediction errors is considered to establish a probability model of uncertain resources. By sampling and simulating the future uncertain resource scenarios through the probability model, refined production simulation is carried out according to the actual microgrid scheduling method, and the probability distribution of the microgrid revenue is obtained by using the Monte Carlo simulation method. The present invention considers the rationality of simulating the uncertain resource scenarios. When calculating the operation cost revenue of the microgrid, the electricity charges for purchasing / selling electricity from / to the public grid take into account the influence of the two-part electricity price and the time-of-use electricity price on the microgrid revenue. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 is a schematic diagram of a probabilistic revenue analysis method applicable to the user-side microgrid of the present invention;

[0032] Figure 2 It is a schematic diagram of a microgrid system in an embodiment of a probability revenue analysis method for a user-side microgrid according to the present invention;

[0033] Figure 3 It is the summer electricity price curve in an embodiment of a probability revenue analysis method for a user-side microgrid according to the present invention;

[0034] Figure 4 It is the load data graph in an embodiment of a probability revenue analysis method for a user-side microgrid according to the present invention;

[0035] Figure 5 It is the photovoltaic output data graph in an embodiment of a probability revenue analysis method for a user-side microgrid according to the present invention;

[0036] Figure 6 It is the correlation coefficient of prediction errors for adjacent time periods of high-level load types in a probability revenue analysis method for a user-side microgrid according to the present invention;

[0037] Figure 7 It is the monthly revenue probability distribution graph (investing 500 kWh / 170 kW energy storage) in a probability revenue analysis method for a user-side microgrid according to the present invention;

[0038] Figure 8 It is the monthly revenue probability distribution graph (investing 750 kWh / 250 kW energy storage) in a probability revenue analysis method for a user-side microgrid according to the present invention. Detailed implementation manners

[0039] The following combines with the Figures 1 to 8 The given embodiments further illustrate the detailed implementation manners of a probability revenue analysis method for a user-side microgrid according to the present invention. The probability revenue analysis method for a user-side microgrid according to the present invention is not limited to the descriptions of the following embodiments.

[0040] A probability revenue analysis method for a user-side microgrid includes the following steps:

[0041] S1: Obtain the historical data and prediction data of each day, cluster the historical curves in the daily historical data to obtain the daily types of load and renewable energy power generation, and select the combinations of daily types of load and renewable energy power generation to form multiple daily typical types;

[0042] S2: Select a daily typical type, and obtain the simulation scenario set of the selected daily typical type according to the historical data and prediction data of each day;

[0043] S3: Repeat step S2 until the simulation scenario sets of all daily typical types are obtained, and then perform step S4;

[0044] S4: Calculate the revenue and the probability of revenue by randomly sampling from the simulation scenario sets of each daily typical type according to the proportion of each daily typical type.

[0045] The present invention discloses a probabilistic revenue analysis method applicable to a user-side microgrid. This method transforms the uncertainty of the revenue of the user-side microgrid into a probability problem for analysis. First, analyze the distribution of uncertain resources (loads, photovoltaic power generation, wind power generation, etc.) in the microgrid on a long time scale, and consider the influence of prediction errors to establish a probability model of uncertain resources. Sample and simulate future uncertain resource scenarios through the probability model, conduct refined production simulation according to the actual microgrid scheduling method, and use the Monte Carlo simulation method to obtain the probability distribution of the microgrid revenue. The present invention considers the rationality of simulating uncertain resource scenarios. When calculating the operating cost revenue of the microgrid, the electricity charges for purchasing / selling electricity from the public grid take into account the impact of the two-part electricity price and time-of-use electricity price on the microgrid revenue.

[0046] A probabilistic revenue analysis method applicable to a user-side microgrid includes the following steps:

[0047] Step S1: Obtain historical data and prediction data of each day from the database. The historical data includes historical curves of daily loads and other influencing factors (in this embodiment, the other influencing factors are renewable energy generation powers, such as photovoltaic and wind power, etc.). And make predictions based on the historical curves of daily loads and renewable energy generation powers, as well as the daily weather type, maximum temperature, and minimum temperature to obtain corresponding prediction data; respectively use the Kmeans clustering method for the historical curves of daily loads and renewable energy generation powers to obtain multiple daily types. The daily types include the proportion situations of their respective loads and renewable energy generation powers. Select two or more from the multiple daily types for free combination to obtain multiple daily typical types and their proportions. The product of the proportions of the daily types selected for free combination is used as the proportion of the daily typical type. The prediction data includes a load prediction curve and a prediction curve of renewable energy generation power.

[0048] Preferably, in step S1, conduct revenue analysis quarterly / monthly. Assume that the daily loads and daily renewable energy generation powers within each quarter or month are independent, and the proportions of each type of load and renewable energy generation power remain unchanged in the same future time period. In this embodiment, the renewable energy generation power includes photovoltaic output and wind power output, and the proportions of each typical type are:

[0049] P(l) = P load (i)·P pv (j)·P wind (k)

[0050] l = 1...n, i = 1...n load , j = 1...n pv , k = 1...n wind

[0051] where n is the number of typical types, n = n load ·n pv ·n wind ; n load , n pv , n wind are respectively the numbers of daily historical curve types of load, photovoltaic output, and wind power output.

[0052] Step S2: Analyze the daily typical types, which specifically includes the following steps:

[0053] Step S2.1: Select one of the daily typical types obtained in Step S1 for analysis. Calculate the absolute prediction error based on the historical load curve, load prediction curve, historical curve of renewable energy generation power, and its prediction curve related to the daily typical type. Use the non - parametric kernel density estimation method to obtain the prediction error probability model, that is, use the non - parametric kernel density estimation method to describe the probability distribution of the prediction error at each moment and the joint probability distribution of the prediction errors at adjacent moments. Without making assumptions about the distribution model, directly input the historical prediction error to obtain the probability distribution model.

[0054] Step S2.2: In the selected daily typical type, randomly sample the historical curve of the load and the historical curve of renewable energy generation power respectively, and use the corresponding prediction curve for day - ahead scheduling.

[0055] Step S2.3: According to the prediction curve and the prediction error probability model, obtain the simulated scenario curve for real - time scheduling and conduct a refined production simulation once.

[0056] In Step S2.3, the refined production simulation of the micro - grid is carried out using the actual scheduling strategy. The refined production simulation is to give the scheduling strategy of the micro - grid, the predicted and actual data of load and photovoltaic, and other given regulations, and simulate the operation of the micro - grid through programming, which belongs to the prior art. The scheduling strategy is preferably divided into two stages: day - ahead plan and real - time scheduling.

[0057] In the day - ahead plan, the prediction results of the load and renewable energy generation power are directly obtained by random sampling in the prediction curve. The specific stage of the day - ahead plan includes the following steps: divide the whole day into several time periods, use the prediction results of the load and renewable energy generation power, and take the minimum operation cost of the whole - day micro - grid as the optimization goal, and model the day - ahead plan as a mixed - integer linear programming problem.

[0058] In the real-time scheduling stage, the whole day is divided into multiple optimization periods, where the interval of each optimization period is less than that of each period divided in the day-ahead plan. In this embodiment, the day-ahead plan stage divides the whole day into 96 periods, and the real-time scheduling stage divides the whole day into 1440 periods. For the next optimization period, the actual values of the current load and the renewable energy generation power are used as the load and the renewable energy generation power for the next period for real-time scheduling. With the goal of minimizing the deviation between each device of the power grid and the day-ahead plan, the real-time scheduling is modeled as a quadratic programming problem.

[0059] In the real-time scheduling stage, the prediction error at each moment is simulated point by point. The simulated values of the load and the renewable energy generation power at each moment are obtained by subtracting the prediction error of each point from the prediction curve. Among them, the prediction error at the first moment is randomly sampled from the corresponding prediction error probability distribution, and the prediction errors at other moments are randomly sampled from the corresponding joint probability distribution of the prediction errors after determining the prediction error at the previous moment.

[0060] Step S2.4: Repeat steps S2.2 - S2.3 for N times until a set of simulated scenarios of the selected daily typical type is obtained. Since random sampling of historical data is performed each time in step S2.2, the corresponding prediction curves may also be different. In step S2.3, the prediction error at each moment is sampled according to the prediction error probability model, and a simulated scenario is obtained based on the prediction curve, that is: simulated value = predicted value - prediction error. Therefore, different simulated scenarios can be obtained by repeating multiple times. Production simulation is carried out according to the simulated scenarios to obtain how the microgrid operates in each period of this scenario, and thus the operating cost of the microgrid is calculated.

[0061] Step S3: After obtaining the set of simulated scenarios of all daily typical types, perform step S4.

[0062] Step S4: Randomly sample according to the proportion of each daily typical type in the set of simulated scenarios of each daily typical type, and calculate the revenue. Preferably, the calculation of the revenue is performed monthly or quarterly. The probability distribution of the monthly revenue is obtained through more than 10,000 simulations of the monthly revenue.

[0063] The specific process is as follows. When simulating the monthly revenue each time, according to the proportion of daily typical types, randomly determine the number of days Nj of each daily typical type in that month, and respectively select Nj results from the simulation scenario sets of the corresponding daily typical types, thus constituting the operating cost of each day in that month. Then calculate the demand cost of that month according to the maximum demand value in the selected daily. The total cost of that month is the sum of the daily operating costs plus the monthly demand cost. Comparing it with the total cost before configuring energy storage (that is, when not considering energy storage, the electricity cost of the load and the monthly demand cost) can obtain the monthly revenue. Probability revenue analysis is essentially an uncertain evaluation of the costs and revenues of a microgrid over a period of time, and has multiple application scenarios, such as the evaluation of the operating cost revenue of a microgrid, the risk decision-making of microgrid equipment investment, and the optimization of microgrid dispatching schemes, etc.

[0064] The method of this application is applicable to microgrids of any type and any capacity ratio, composed of controllable power sources, uncontrollable power sources and energy storage devices, under various operating modes. When analyzing the revenue, it is necessary to specify the microgrid configuration parameters, operating mode and electricity price calculation method, and also the load, actual historical daily curves and predicted curves of renewable energy generation power of the microgrid for a certain period of history. The costs of the user-side microgrid include the investment and operation and maintenance costs of the converted equipment (energy storage system, controllable power source, uncontrollable power source, control system, etc.). The revenues of the user-side microgrid include but are not limited to the savings in two-part electricity charges (basic electricity charge and electroplating electricity charge), and the revenue from selling surplus electricity back to the grid.

[0065] Combined with Figures 1-4 and Table 1-3, a specific calculation process is provided as follows:

[0066] The microgrid system of a certain park is as Figure 2 shown. The basic parameters of the microgrid system are shown in Table 1, and the electricity price situation of the park microgrid in summer is shown in Table 2.

[0067] The basic parameters of the microgrid system are shown in Table 1:

[0068] Table 1 Basic parameters of the microgrid system

[0069]

[0070]

[0071] From such as Figure 4 、 5Obtain the historical data and predicted data of the system's load and photovoltaic power generation for a certain month from the database shown. Divide the time of a day into 96 points, that is, the data time interval is 15 minutes. On this basis, conduct a probability profit analysis for this month. Use the K-means clustering algorithm to cluster the historical curves of the load and photovoltaic output for each day of this month respectively. According to the clustering results, divide the load into three categories: low-level load, medium-level load, and high-level load, and divide the photovoltaic output into two categories: strong light resources and weak light resources. Select two of them for free combination to obtain 6 typical types. Multiply the proportions of the two types in the free combination to obtain the proportion of the daily typical type, and obtain the proportions of 6 daily typical types according to the proportion combinations of the load and photovoltaic output respectively (see Table 3).

[0072] Table 3 Proportions of each daily typical type

[0073]

[0074] Taking the high-level load type as an example, calculate the Pearson correlation coefficient of the prediction errors between adjacent time periods in the high-level load type to verify the correlation of the prediction errors between adjacent time periods. The detailed steps are as follows:

[0075] a) First, use the historical load data and load prediction data to calculate the absolute prediction error of each time period of this type of load for each day. The absolute prediction error calculation formula is:

[0076] ΔP = P′ - P

[0077] where ΔP is the absolute prediction error, P′ is the load prediction value, and P is the load true value.

[0078] b) Set the prediction errors of the same time period of all historical days as a group of random variables. In this embodiment, divide the data of a day into 96 points, that is, 96 groups of random variables can be obtained. Calculate the Pearson correlation coefficient of the prediction errors between adjacent moments according to the following formula:

[0079]

[0080] where ΔP i and ΔP i+1 are the prediction error random variables at the i-th and i+1-th moments respectively, and r i is the correlation coefficient between the prediction errors at the i-th and i+1-th moments.

[0081] The calculation result is as Figure 6 shown.

[0082] Under the high-level load type, the correlation coefficient of the prediction errors in adjacent time periods is greater than 0.7 most of the time, and the mean value is about 0.79. There is no unified standard for analyzing the correlation using the Pearson correlation coefficient. Generally, when the absolute value of the Pearson correlation coefficient is greater than 0.6 or 0.7, it can be considered that the correlation between two variables is strong. Therefore, under the high-level load type, the correlation of the prediction errors in adjacent time periods is strong. Similar results also exist for other types of load and PV prediction errors. Therefore, it is reasonable to sample the prediction errors by modeling using the joint probability distribution of the prediction errors in adjacent time periods, and the non-parametric kernel density estimation method is used to obtain the joint probability distribution of the prediction errors in each adjacent time period.

[0083] According to the prediction error random variables at each moment, the non-parametric kernel density estimation theory is used to obtain the probability distribution of the prediction errors at each time of various types of load and PV output power, as well as the joint probability distribution of the prediction errors at adjacent times.

[0084] Suppose \(x_1,x_2,\cdots,x\) n is a sample of a d-dimensional random vector obtained from a distribution described by an unknown probability density function \(f\). The kernel density estimation is defined as:

[0085]

[0086] where \(H\) is a symmetric positive definite \(d\times d\) bandwidth matrix; \(K\) is the kernel function, which is a symmetric multi-dimensional probability density function.

[0087] Currently, most programming languages have extension libraries for kernel density estimation. In this application, Kernel Density in the Julia language is used. Only by inputting the prediction error random variables at each moment can the results of kernel density estimation be obtained. The economic dispatch of the microgrid in this park adopts the day-ahead plan plus real-time dispatch method. The simulation method for one day in a month is as follows: According to the proportion of each daily typical type, randomly determine the typical type of this day, and randomly select the load and PV prediction curves under this typical type from the simulation scenario set of this typical type for the day-ahead plan. For example, if the typical type is I and the proportion of typical type I is 20%, then in the whole month, the probability that the daily type of each day is typical type I is 20%. The corresponding of this typical type I is load type I + PV type I. Then, it is necessary to select a load prediction curve of a certain day belonging to load type I and a PV prediction curve of a certain day belonging to PV type I from the historical data; using the prediction error distribution obtained above, randomly sample at each moment to determine the prediction error at each moment, and thus obtain the simulated load and PV curves based on the prediction curves for real-time dispatch. From this, the operating conditions of the microgrid such as the PCC point power, energy storage charge and discharge power, and diesel generator operating power at each time period of this day can be obtained.

[0088] The monthly operating cost of the microgrid includes the monthly demand electricity cost DC and the daily electricity consumption cost C grid (including the income from selling surplus electricity to the grid), and the power supply cost C of the diesel generator G . The calculation method of the monthly demand electricity cost is the demand electricity price multiplied by the actual maximum demand of the month. The actual maximum demand of the month refers to the maximum value of the power (15-minute average value) in a month. The monthly demand electricity cost, the daily electricity consumption cost, the power supply cost of the diesel generator and the calculation formula are as follows:

[0089]

[0090] where c DC is the rate of the monthly demand cost; is the maximum value of the power purchased by the microgrid from the public grid in this month; is the power purchased by the microgrid from the public grid at time t; is the electricity purchase rate corresponding to the time-of-use electricity price; is the power of the microgrid's surplus electricity sold to the grid at time t; is the surplus electricity selling rate corresponding to the surplus electricity selling price; is the operating power of the diesel generator at time t; is the rate of the diesel generator's power supply.

[0091] According to the simulation situation of each day and the actual maximum demand, the operating cost of the microgrid in this month can be calculated. When calculating the daily electricity consumption cost, the power of purchasing / selling electricity from / to the main grid in each period obtained from the production simulation is multiplied by the corresponding electricity purchase / selling rate in that period to obtain the electricity consumption cost of each period. This calculation process takes into account the time-of-use electricity price, that is, the electricity price is different in different periods; when calculating the monthly demand cost, according to the results of the daily production simulation of the whole month, the maximum power of purchasing electricity from the main grid in this month can be obtained, and this power is the maximum demand of this month. The maximum demand multiplied by the rate of the demand cost is the demand cost. For example, if the maximum power of the microgrid purchasing electricity from the main grid in August is 500 kW and the rate of the monthly demand cost is 42 yuan / kW, then the monthly demand cost is 500 * 42 = 21000 yuan. By comparing the operating cost of the microgrid before the energy storage is put into use, the income brought by the energy storage in this month can be obtained. Using the Monte Carlo simulation method, a large number of simulations (more than 10000 times) are carried out for this month, and the probability distribution of the monthly income is obtained by using kernel density estimation according to the results of each simulation.

[0092] Suppose the microgrid invests in a 500 kWh / 170 kW energy storage system with an investment cost of 1 million yuan, then the probability density function and cumulative distribution function of the monthly income obtained according to the above method are as Figure 7 shown ( Figure 7In it, the upper part is the probability density function of monthly income, and the lower part is the cumulative distribution function of income. The points on the curve are such that the ordinate is the probability that the monthly income exceeds the abscissa value).

[0093] From the analysis results, after the microgrid invests in a 500 kWh / 170 kW energy storage system, the probability that the income exceeds 16,664 yuan is 90%, the probability that the income exceeds 19,966 yuan is 50%, and the probability that the income exceeds 22,096 yuan is 1%. Therefore, in the risk decision-making of energy storage investment, we can consider that: 16,664 yuan is a conservative estimate of the monthly income, the median of the expected monthly income is about 19,966 yuan, and the maximum value that the monthly income may reach is about 22,096 yuan.

[0094] Suppose the microgrid invests in a 750 kWh / 250 kW energy storage system, and the investment cost is 1.5 million yuan. The probability distribution result of the monthly income is as Figure 8 shown. At this time, the conservative estimate value of the monthly income is 22,744 yuan, the median of the expected monthly income is about 25,150 yuan, and the maximum value that the monthly income may reach is about 29,393 yuan. It can be seen that increasing the scale of energy storage investment can effectively improve the electricity cost savings income brought by energy storage, but the cost required to invest in the energy storage system will also increase. Therefore, it is necessary to weigh between the investment cost and the income according to the actual goals and needs.

[0095] Assume that the goal of this project is to recover the cost as soon as possible under the conservative estimate of income. Calculate the conservative estimate value of the monthly income of the whole year in the above way. The conservative estimate of the annual income of the microgrid investing in a 500 kWh / 170 kW energy storage system is 174,000 yuan, and the conservative estimate of the annual income of the microgrid investing in a 750 kWh / 250 kW energy storage system is 241,000 yuan. Then, without considering the discount rate, based on the input costs of the known 500 kWh / 170 kW and 750 kWh / 250 kW energy storage systems, the cost recovery periods are 5.75 years and 6.22 years respectively. Therefore, it is more appropriate to choose 500 kWh / 170 kW.

[0096] The load renewable energy in the microgrid in this application is uncertain, so the income is an uncertain value. When calculating the income of the microgrid, a specific microgrid needs to be selected, and a specific scheduling strategy also needs to be selected according to the microgrid. And it is not limited to these two types of day-ahead plans and real-time scheduling. As long as the microgrid scheduling depends on day-ahead prediction, the method in this patent can be used for probability income analysis. In addition, the day-ahead plan is a mixed-integer linear programming problem, and the real-time scheduling is a quadratic programming problem, which are prior arts.

[0097] The present invention also provides a computer, comprising a processor and a memory. The memory stores computer instructions for implementing the probability revenue analysis method applicable to the user-side microgrid of the present invention, and the processor executes the computer instructions stored in the memory to perform the steps in the probability revenue analysis method for the user-side microgrid.

[0098] A computer-readable storage medium storing a computer program, which, when executed by one or more processors, causes the one or more processors to perform the steps in the probability revenue analysis method applicable to the user-side microgrid of the present invention.

[0099] The above content is a further detailed description of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is only limited to these descriptions. For those of ordinary skill in the technical field to which the present invention pertains, without departing from the concept of the present invention, several simple deductions or substitutions can be made, and all should be regarded as belonging to the protection scope of the present invention.

Claims

1. A probabilistic benefit analysis method applicable to user-side microgrids, characterized in that: It includes the following steps: S1: Obtain the historical data and predicted data of each day, cluster the historical curves in the daily historical data to obtain the daily types of load and renewable energy power generation, and select the combinations of the daily types of load and renewable energy power generation to form multiple daily typical types; The historical data includes the historical load curve of each day and the historical curves of at least one type of renewable energy power generation. The predicted data includes the load prediction curve and the predicted curve of renewable energy power generation. Cluster the historical load curve and the historical curves of renewable energy power generation respectively to obtain multiple daily types and the proportions of the daily types. Select one daily type of load and its proportion and at least one daily type of renewable energy power generation and its proportion for free combination to obtain multiple daily typical types and their proportions; The renewable energy power generation includes photovoltaic output and wind power output. If the load, photovoltaic output, and wind power output of each day are independent of each other and the proportions of each type of load, photovoltaic output, and wind power output remain unchanged in a future period of time, the proportions of each daily typical type are: P(l) = P load (i)·P pv (j)·P wind (k) l = 1...n, i = 1...n load , j = 1...n pv , k = 1...n wind where n is the number of daily typical types, n = n load ·n pv ·n wind ; n load n pv n wind are the numbers of daily historical curve types of load, photovoltaic output, and wind power output respectively; S2: Select a daily typical type, and obtain the simulation scenario set of the selected daily typical type according to the historical data and predicted data of each day, including the following steps, S2.1: Select one of the multiple daily typical types, calculate the absolute prediction error according to the historical load curve, load prediction curve, historical curve of renewable energy power generation, and its prediction curve related to the daily typical type, and obtain the prediction error probability model; S2.2: In the selected daily typical type, randomly sample the historical curve of the load and the historical curve of renewable energy power generation respectively, and perform day-ahead scheduling using the prediction curve corresponding to the sampled sample; S2.3: According to the prediction curve and the prediction error probability model, obtain the simulation scenario curve for real-time scheduling and perform refined production simulation; S2.4: Repeat steps S2.2 - S2.3 for N times until the simulation scenario set of the selected daily typical type is obtained; S3: Repeat step S2 until the simulation scenario sets of all daily typical types are obtained, and perform step S4; S4: According to the proportions of each daily typical type, randomly sample and calculate the revenue and the probability of the revenue in the simulation scenario sets of each daily typical type.

2. The probabilistic benefit analysis method for a user-side microgrid according to claim 1, characterized in that: In step S2, the non-parametric kernel density estimation method is used to describe the prediction error probability distribution at each moment and the joint probability distribution of the prediction errors between adjacent moments.

3. The probabilistic benefit analysis method for the user-side microgrid according to claim 1, characterized in that: In step S2.3, the actual scheduling strategy is used for refined production simulation. The scheduling strategy includes two stages: day-ahead plan and real-time scheduling; In the day-ahead plan stage, the whole day is divided into several time periods, and random sampling is performed in the load prediction curve and the predicted curve of renewable energy power generation obtained in step S1. With the minimum operation cost of the microgrid throughout the day as the optimization goal, the day-ahead plan is modeled as a mixed-integer linear programming problem; In the real-time scheduling stage, for the next optimization period, using the current load and renewable energy generation power, with the minimum deviation between the microgrid devices and the day-ahead plan as the optimization goal, the real-time scheduling is modeled as a quadratic programming problem.

4. A probability revenue analysis method applicable to a user-side microgrid according to claim 3, characterized in that: In the real-time scheduling stage, the prediction error at each moment is simulated point by point. The simulated load value and the simulated renewable energy generation power value at each moment are obtained by subtracting the prediction error at each point from the load prediction curve and the renewable energy generation power prediction curve. Among them, the prediction error at the first moment is randomly sampled from the corresponding prediction error probability distribution, and the prediction errors at other moments are randomly sampled from the corresponding joint probability distribution of prediction errors after determining the prediction error at the previous moment.

5. A probabilistic benefit analysis method applicable to the user-side microgrid according to claim 1, characterized in that: In step S4, according to the proportions of each daily typical type, randomly determine the daily typical type of a certain day, and randomly select the load and renewable energy generation power prediction curves under this daily typical type from the simulated scenario set of this daily typical type for the day-ahead plan. Using the obtained error prediction distribution model, randomly sample at each moment to determine the prediction error at each moment, and obtain the simulated load and renewable energy generation power curves based on the prediction curves for implementing the scheduling; thus, obtain the operation situation of the microgrid on a certain day, and calculate the microgrid revenue in combination with the daily microgrid operation cost.

6. The probabilistic benefit analysis method for the user-side microgrid according to claim 1, wherein: In step S4, to calculate the revenue, it is necessary to give the microgrid configuration parameters, operation mode, electricity price calculation method, and the historical curves and prediction curves of the load and renewable energy generation of the microgrid for a certain period of time.

7. A probabilistic benefit analysis method applicable to the user-side microgrid according to claim 1, characterized in that: In step S4, when calculating the revenue, it is necessary to consider the cost of the user-side microgrid. The cost of the user-side microgrid includes the converted equipment investment and operation and maintenance costs; the revenue of the user-side microgrid includes the revenue saved from the two-part electricity tariff and the revenue from selling surplus electricity back to the grid.

8. A readable storage medium, characterized in that: A computer program is stored, and when the computer program is executed by one or more processors, the one or more processors are caused to execute the steps in the probability revenue analysis method for the user-side microgrid according to any one of claims 1-7.

Images