CN108321801A - Method and system for making day-ahead power generation plan of energy base system - Google Patents

Abstract

The present invention relates to a method and system for formulating a day-ahead power generation plan of an energy base system, which is characterized in that it includes the following steps: 1) Establishing an optimization of a day-ahead power generation plan of an energy base system including a DC adjustment method according to various prediction information of wind power obtained in advance Model; 2) Formulate a number of application examples containing different types of wind power prediction information, as input variables of the energy base system day-ahead power generation planning model, and determine Optimal day-ahead generation planning for energy base systems. The method of the invention can integrate various wind power prediction information into the optimization model for comparative analysis and verification, and at the same time considers the means of direct current regulation, and can obtain better economic performance under the premise of ensuring the reasonable operation of the system. Therefore, the present invention can be widely used in making the day-ahead power generation plan of the energy base system.

Description

Method and system for formulating day-ahead power generation plan for energy base system

technical field

The present invention relates to the field of making day-ahead power generation plans for electric power systems, in particular to a method and system for making day-ahead power generation plans for energy base systems that take wind power uncertainty into consideration, and is mainly used for making day-ahead power generation plans during the operation of energy base systems.

Background technique

In the current operation mode of wind-fired bundled DC transmission, most of the supporting thermal power and DC operation only support wind power consumption by roughly retaining system backup methods, and the initial design of using thermal power to stabilize the volatility of wind power has not yet been fully realized. Considering the optimal dispatch of wind power, thermal power and DC operation of the UHV DC sending end energy base system, the key is to solve two main problems: one is to make reasonable use of wind power forecast information; the other is to build an optimal dispatch model to coordinate wind power output and thermal power generation cost and the relationship between DC operating costs. The commonly used deterministic unit combination model with reserve constraints uses the given load curve to represent the load fluctuation in the future dispatching period. As for the UHVDC sending-end energy base system, the traditional load curve becomes a DC with limited operating conditions. Therefore, how to reasonably use the wind power prediction information in the current unit combination, consider the influence of wind power uncertainty, and take into account The regulation capacity of the limited DC is a problem faced in the practice of realizing UHV DC external wind power consumption demand.

Contents of the invention

The wind power prediction power in the existing power generation planning method is generally only obtained by one or two methods of deterministic prediction, interval prediction, and scene prediction, and there is a lack of comprehensive application and analysis of wind power prediction information obtained by these prediction methods. Comparative analysis; the current power generation planning methods are mostly based on the joint output of wind and thermal power to balance the load demand at the receiving end, and there is a lack of research on the coordination and optimization of wind and thermal power output in energy bases to meet the requirements of DC output capacity. The purpose of the present invention is to provide a method and system for formulating a day-ahead power generation plan for an energy base system that takes wind power uncertainty into account, utilizes various forecast information of wind power to formulate a day-ahead power generation plan, and considers DC regulation costs at the same time, improving the energy base The wind power consumption capacity of the system is more in line with engineering reality.

In order to achieve the above object, the present invention adopts the following technical solutions: a method for formulating a day-ahead power generation plan of an energy base system, which is characterized in that it includes the following steps: 1) according to the pre-acquired wind power prediction information, establish an optimization of the energy base system day-ahead power generation plan Model; 2) Formulate a number of application examples containing different types of wind power prediction information, as input variables of the energy base system day-ahead power generation planning model, and determine Optimal day-ahead generation planning for energy base systems.

In the step 1), the optimization model of the energy base system’s day-ahead power generation plan includes an objective function and constraint conditions, and the objective function includes an objective function based on point prediction information and an objective function based on scene prediction information; the constraint conditions include Power balance constraints, reserve capacity constraints, unit output constraints, unit ramp rate constraints, unit minimum start-up time constraints, unit minimum downtime constraints and abandoned air volume constraints.

The objective function based on point prediction information is:

In the formula, is the power generation cost of the thermal power unit, is the start-up cost of the thermal power unit, is the downtime cost of the thermal power unit, is the regulation cost of DC, is the cost of wind curtailment, the subscript i represents the i-th thermal power unit, and the subscript t represents the t-th period.

The objective function based on scene prediction information is:

In the formula, is the power generation cost of the thermal power unit, is the start-up cost of the thermal power unit, is the downtime cost of the thermal power unit, is the regulation cost of DC, is the wind curtailment cost, subscript i represents the i-th thermal power unit, subscript t represents the t-th period, subscript j represents the j-th scenario, T is the total number of time periods, N is the total number of thermal power units, S is the total number of scenarios, p _j is the probability of scenario j.

The formulas for calculating the constraints are as follows:

① Power balance constraints:

In the formula: p _it is the output of unit i in period t, is the predicted value of wind power in period t, is the curtailed wind power due to power balance constraints, and is the planned value and adjusted value of the DC power in the period t;

② Reserve capacity constraints:

In the formula: _upit and _dnit are the upspin reserve capacity and downspin reserve capacity of unit i in period t, respectively, and are the upper limit and lower limit of the wind power prediction interval, r _resup and r _resdn are the ratios of up and down reserve capacity, r _ui and r _di are the up and down ramp rate limits of generator i respectively, p _imax and p _imin are generator i t _r is the operating time of the reserve capacity required by the system, α is the interval prediction reserve coefficient, when the optimization model adopts the wind power interval prediction information, α=1, otherwise α=0; z _it is unit i at t The state variable of the period, the value is 0 or 1, when z _it is 1, it means that the unit i starts to run in the period t; when z _it is 0, it means that the unit i shuts down in the period t;

③Unit output constraints:

④ Unit climbing rate constraints:

In the formula, p _it is the output of unit i in period t; p _i(t-1) is the output of unit i in period t-1; Δt is the time interval;

⑤ Constraints on the minimum start-up time of the unit:

In the formula: UT _i and They are the minimum start-up running time of unit i and the start-up time of the initial period, Z _i0 is the operating state of unit i in the initial period, Z _i0 = 1 means that the unit is in the start-up state at this time, Z _i0 = 0 means that the unit is in Off state;

⑥Minimum downtime constraint:

In the formula: DT _i and are the minimum downtime of unit i and the time it has been down in the initial period, respectively;

⑦ Abandoned air volume constraints:

In the formula: is the curtailed wind power due to power balance constraints, is the curtailed wind power generated due to insufficient down-regulation reserve capacity.

In the step 1), the forecast information of wind power includes deterministic forecast, interval forecast and scene forecast information of wind power.

The calculation method of the scenario prediction information of wind power comprises the following steps: firstly, according to the historical data of the actual output power of the wind farm and the predicted power, the historical data of the prediction error of the wind power is obtained, and an initial wind power error sequence scenario is generated; secondly , based on the average value clustering method, the daily variation dimension scene optimization method, clustering the initial wind power error sequence scene in hours, and generating a representative scene set that can reflect the statistical characteristics of wind power power in each time period of the day; finally, based on the taboo The hourly change dimension scene optimization method of the search method uses the representative scene sets of each time period of the day to select a scene in each time period for connection to form a wind power prediction error sequence. After multiple iterations, similar scene sequences are eliminated to obtain the final Scenario prediction value sequence of wind power.

In the step 2), the method for obtaining the optimal day-ahead power generation plan includes the following steps: 2.1) formulate several application examples that include different types of wind power prediction information; 2.2) use each application example as an input variable, and input it to In the day-ahead power generation plan optimization model, the operation index results of each application example are calculated; 2.3) The results of different operation indicators are compared and analyzed, and the optimal application example is determined according to the actual operation indicators of the energy base system. Optimal day-ahead generation planning for energy base systems.

A day-ahead power generation plan formulation system for an energy base system applicable to the method, characterized in that it includes an optimization model building block and an optimization model calculation module; the optimization model building block is used to obtain wind power prediction information based on , to establish an optimization model of the energy base system's day-ahead power generation plan; the optimization model calculation module is used to formulate several application examples including the use of different types of wind power prediction information, as input variables of the energy base system day-ahead power generation plan model, according to the obtained The comparison results of the operation indicators of the energy base system and the operation indicators focused on determine the optimal day-ahead power generation plan of the energy base system.

The optimization model construction module includes a prediction information acquisition module, an objective function construction module and a constraint condition construction module; the prediction information acquisition module is used to calculate the deterministic prediction, interval prediction and scene prediction information of wind power; the objective function construction The module is used to construct an objective function according to different types of wind power prediction information; the constraint condition building module is used to establish a related constraint function of the objective function; the optimization model calculation module includes a calculation example formulation module, an optimization model calculation module and an operating index comparison module; the calculation example formulating module is used to formulate application examples containing different types of wind power prediction information; the optimization model calculation module is used to calculate and obtain operation index results according to different application examples; the operation index comparison results It is used to compare and analyze the obtained results of different operating indicators, and obtain the optimal application example as the optimal day-ahead power generation plan.

Due to the adoption of the above technical solutions, the present invention has the following advantages: 1. The present invention utilizes various prediction information of wind power to establish a day-ahead power generation plan optimization model, and then formulates a day-ahead power generation plan. Considering factors are more comprehensive, and can be based on actual The wind power forecasting situation or the comparison of the indicators focused on are used for selection. 2. The day-ahead power generation plan optimization model established by the present invention takes DC regulation costs into consideration, and the obtained day-ahead power generation plan is more in line with engineering reality. Therefore, the present invention can be widely used in making day-ahead power generation plans of energy base systems. 3. When the present invention calculates the scene prediction information of the wind farm, it adopts the combination of the daily change dimension scene analysis method based on the mean value clustering method and the hour dimension scene optimization method based on the tabu search method, which greatly reduces the number of representative scenes. At the same time, the accuracy of the reduction results is taken into account, and the statistical law of wind power power is simulated with a small number of scenarios, which provides important basic information for the operation and planning of the power system under the background of large-scale wind power grid integration. Therefore, the present invention can be widely used in the formulation of energy base system power generation plan.

Description of drawings

Figure 1 is a flow chart of daily-variation dimension scenario optimization based on the mean value clustering method;

Figure 2 is a flow chart of scene optimization based on the tabu search method for the hour-changing dimension;

Figure 3 is the comparison between the predicted value of wind power and the actual value;

Fig. 4 is a certain power interval prediction error fitting probability density function;

Figure 5 is the 90% probability interval of wind power;

Figure 6(a) to Figure 6(c) are demonstration diagrams of the general idea of the two-dimensional optimization method for wind power sequence;

Figure 7 is the best representative scene for each time period;

Figure 8 is the probability corresponding to the optimal scene in each time period;

Figure 9 is a comparison of the mean value of the original scene and the optimized scene;

Figure 10 shows the variance comparison between the original scene and the optimized scene;

Fig. 11 is fitness function trend curve;

Figure 12 shows the corresponding probability of each scene;

Figure 13 is a representative scene of the prediction error sequence after the optimization of the hour change dimension;

Figure 14 is a comparison chart of the planned operation value before DC adjustment and the planned operation value after DC adjustment.

Detailed ways

The present invention will be described in detail below in conjunction with the accompanying drawings and embodiments.

The present invention first utilizes the BP neural network algorithm optimized by the PSO algorithm (i.e. the PSO-BP algorithm) to predict the wind power with certainty, and on this basis, uses a non-parametric regression model to perform interval prediction of the wind power, and adopts the method of scene analysis to consider wind power. Uncertainty of output, and use the two-dimensional scene optimization method based on the daily change dimension and hour change dimension to reduce the scene, so as to ensure the richness of the optimized scene and the accuracy of the final result. A day-ahead power generation planning model that includes wind power forecast information and DC regulation costs is constructed to improve the wind power absorptive capacity of the energy base system and provide suggestions for the formulation of the day-ahead power generation plan of the energy base system.

A method for formulating a day-ahead power generation plan for an energy base system provided by the present invention includes the following steps:

1) Obtain various forecast information of wind power, including deterministic forecast, interval forecast and scene forecast information of wind power.

The calculation method of various forecast information of wind power includes the following steps:

1.1) Deterministic prediction of the wind power of the energy base system is carried out to obtain the deterministic prediction information of the wind power.

The deterministic prediction information of wind power is calculated by PSO-BP (Particle Swarm Optimization Backpropagation) neural network algorithm. The PSO-BP neural network algorithm combines the particle swarm algorithm with the BP neural network, that is, the particle swarm algorithm is used in the training of the BP neural network, the connection weights and various thresholds in the BP neural network are optimized, and the trained neural network is used Deterministic forecasting of wind power. When the PSO-BP neural network algorithm is used to predict wind power deterministically, the input variables are the cosine and sine values of wind speed and wind direction angle, and the output variable is wind power. Since this algorithm is a prior art, the present invention will not repeat it here.

1.2) Using the kernel density estimation method in non-parametric estimation, the interval prediction of the wind power of the energy base system is carried out, and the interval prediction information of the wind power is obtained.

A method for interval forecasting of wind power comprises the following steps:

1.2.1) According to the magnitude of the predicted value in the historical data of the predicted power of the wind farm, the range is divided, and the predicted power is divided into several power ranges.

The specific operation method is: calculate the average value of the wind power forecast value in each scheduling period, convert the forecast value of the wind power point at 96 points throughout the day into the forecast value of 24 points, and then divide it into equal intervals. Assuming that the maximum value and minimum value of the wind power prediction value are P _max and P _min respectively, and the length of the power interval is L, then the number N of divided power intervals is:

N=(P _max -P _min )/L+1 (1)

The i-th power interval Z _i is:

Z _i =[P _min +(i-1)L,P _min +iL]i=1,2,L,N (2)

1.2.2) Merge and divide the power intervals in step 1.2.1) according to the preset sample point threshold, and after obtaining the final power intervals, establish a wind power prediction error distribution model for each power interval.

If the number of sample points in the obtained power interval does not meet the preset sample point threshold, it will be merged with the power interval adjacent to the power interval and the number of sample points also does not meet the sample point threshold, until the merged new The number of sample points in the power interval meets the threshold of sample points. Among them, the sample point threshold refers to half of the idealized average sample point, and the idealized average sample point is equal to the total number of samples divided by the number of intervals. According to the obtained new power range, the wind power prediction error distribution model of each power range is established.

1.2.3) Based on the established wind power forecasting error distribution model in each power interval, the probability density function of the forecasting error samples in each power interval is calculated by using the non-parametric kernel density estimation method, and the corresponding error probability density curve is obtained.

Assuming that a certain prediction error is e, its probability density function is:

In the formula, N is the total number of samples, h is the bandwidth coefficient, e is the random variable of the prediction error, e _i is the i-th prediction error sample; K(u) is the probability density function, when the Gaussian kernel function is used, K(u )for:

1.2.4) For each deterministic prediction value of wind power, determine the power interval to which it belongs, and search for the error probability density curve corresponding to the power interval, and convert to obtain the power probability density curve.

1.2.5) For each wind power deterministic forecast value, calculate the power interval including the value at the set confidence level, select the power interval with the smallest interval length as the confidence interval of the predicted value, and record the corresponding confidence interval boundary.

1.2.6) The upper and lower bounds of the confidence intervals corresponding to all predicted values are differentially fitted to obtain an interval curve containing all predicted values with expected probability, and then the interval predicted value of wind power is obtained.

1.3) Using the scenario analysis method to model the uncertainty of wind power, using the daily variation dimension scenario optimization method based on the mean value clustering method and the hour dimension scenario optimization method based on the tabu search method to optimize the wind power sequence scenario, we get Scenario forecast value of wind power.

The method of optimizing the wind power sequence by adopting the two-dimensional scene optimization method to obtain the scene prediction value of the wind power comprises the following steps:

1.3.1) According to the historical data of the actual output power and predicted power of the wind farm, the historical data of the forecast error of wind power power is obtained, and the initial wind power error sequence scene is generated.

1.3.2) The daily variation dimension scene optimization method based on the mean value clustering method clusters the initial wind power error sequence scene in step 1.3.1) with hours as the time period, and generates wind power statistics that can reflect the statistical characteristics of wind power in each time period of the day. Represents a collection of scenes.

As shown in Figure 1, the daily variation dimension scenario optimization method based on the mean value clustering method, when clustering the initial wind power error sequence scenario, includes the following steps:

①Determine the number of reserved scenes G _s in each period, randomly select G _s scenes from all the scenes as the core, and generate the initial core scene set:

② Establish the remaining scene collection (s′=1,2,LN _s -G _s ), and calculate the distance from each remaining scene to the core scene

③According to the calculated distance d _s,s′ , classify the remaining scenes into the core scenes closest to it, and obtain the classified cluster set Cluster={C _i }, i=1,2,LG _s , In each similar scene set, select the scene with the smallest sum of distances from other scenes as the new core;

④Repeat steps ②～③ until the core does not change any more and the scene reduction ends, and the probability of each scene is calculated as the sum of the probabilities of all the scenes in the cluster where the scene is located;

⑤ The scenarios and their probabilities in the obtained core scenario set are representative scenario sets that can reflect the statistical characteristics of wind power at each time period of the day.

1.3.3) Based on the tabu search method, the hourly change dimension scene optimization method uses the representative scene sets of each time period formed in step 1.3.2), selects a scene in each time period for connection, and forms a wind power prediction error sequence, After multiple iterations, similar scene sequences are eliminated to obtain the final sequence of wind power scene prediction values.

As shown in Figure 2, the hour-changing dimension scene optimization method based on the tabu search method includes the following steps:

①Given the optimal scene to solve the number of scenes Q Generate an initial feasible solution Q ₀ ;

②Empty the taboo table, calculate the fitness function f ₀ of the initial feasible solution Q ₀ , the number of iterations k _iter =0, and set the termination condition ε;

③Calculate the fitness function value of the current solution of the k _iter iteration and its neighborhood solutions;

④ Compare the fitness function value of the current solution and its neighborhood solutions, and take the maximum value as the fitness function value of the optimal solution of the k _iter iteration, namely The corresponding solution is the optimal solution

⑤ Determine whether the termination condition is satisfied, that is, Whether it is satisfied, if it is satisfied, stop searching and output optimization results Otherwise, set the number of iterations k _iter = k _iter +1, and add all scenarios of all solutions except the optimal solution to the taboo list, repeat ③～⑤ until the termination condition is satisfied, and then stop the search;

⑥ All the scenarios obtained from the optimal solution are the sequence of final wind power scenario prediction values.

2) According to the various prediction information of wind power obtained, an optimization model of the energy base system's day-ahead power generation plan including DC regulation means is established.

In the established optimization model of day-ahead power generation plan, the objective function includes power generation cost, start-up cost, shutdown cost, DC regulation cost and wind curtailment cost of thermal power units. According to the different types of wind power forecast information, the objective function of the day-ahead power generation planning optimization model of the energy base system is determined:

When predicting information based on points, the objective function is:

When predicting information based on the scene, the objective function is:

In the formula: is the power generation cost of the thermal power unit, is the start-up cost of the thermal power unit, is the downtime cost of the thermal power unit, is the regulation cost of DC, is the wind curtailment cost, subscript i represents the i-th thermal power unit, subscript t represents the t-th period, subscript j represents the j-th scenario, T is the total number of time periods, N is the total number of thermal power units, S is the total number of scenarios, p _j is the probability of scenario j.

Constraints include power balance constraints, reserve capacity constraints, unit output constraints, unit ramp rate constraints, unit minimum start-up time constraints, unit minimum downtime constraints and abandoned air volume constraints. The details are as follows:

① Power balance constraints:

In the formula: p _it is the output of unit i in period t, is the predicted value of wind power in period t, is the curtailed wind power due to power balance constraints, and is the planned value and adjusted value of the DC power in the period t.

② Reserve capacity constraints:

In the formula: _upit and _dnit are the upspin reserve capacity and downspin reserve capacity of unit i in period t, respectively, and are the upper and lower limits of the wind power prediction interval, r _resup and r _resdn are the ratios of up and down reserve capacity, r _ui and r _di are the up and down ramp rate limits of generator i respectively, p _imax and p _imin are generator i t _r is the operating time of the reserve capacity required by the system, α is the interval prediction reserve coefficient, when the optimization model adopts the wind power interval prediction information, α=1, otherwise α=0; z _it is unit i in t period The state variable of , takes a value of 0 or 1. When z _it is 1, it means that the unit i starts to run in the period t; when z _it is 0, it means that the unit i shuts down in the period t.

③Unit output constraints:

④ Unit climbing rate constraints:

In the formula, p _it is the output of unit i in period t; p _i(t-1) is the output of unit i in period t-1; Δt is the time interval. The units of the up and down ramp rates r _ui and r _di (the subscript i represents the i-th unit) are generally WM/min, so △t is 60.

⑤ Constraints on the minimum start-up time of the unit:

In the formula: UT _i and They are the minimum start-up running time of unit i and the start-up time of the initial period, Z _i0 is the operating state of unit i in the initial period, Z _i0 = 1 means that the unit is in the start-up state at this time, Z _i0 = 0 means that the unit is in Off state.

⑥Minimum downtime constraint:

In the formula: DT _i and are the minimum downtime of unit i and the time it has been down during the initial period, respectively.

⑦ Abandoned air volume constraints:

In the formula: is the curtailed wind power due to power balance constraints, is the curtailed wind power generated due to insufficient down-regulation reserve capacity.

3) Formulate a number of application examples based on the use of different types of wind power prediction information, as the input variables of the energy base system's day-ahead power generation planning model, and determine the energy base based on the comparison results of the energy base system's operating indicators and the operating indicators that are focused on The optimal day-ahead power generation plan for the system.

3.1) Formulate a number of application examples including the use of different types of wind power prediction information;

3.2) Take each application example as an input variable, input it into the day-ahead power generation plan optimization model, and calculate the operation index results of each application example;

3.3) Compare and analyze the obtained results of different operating indicators, and determine the optimal application example as the optimal day-ahead power generation plan of the energy base system according to the actual operating indicators of the energy base system.

The present invention will be further described in detail below in conjunction with specific embodiments. The present invention uses relevant data of a certain wind farm in Germany in the past year (June 1, 2015 to May 30, 2016) for calculation to verify the effectiveness of the present invention.

1) Obtain various forecast information of wind power, including deterministic forecast, interval forecast and scene forecast information of wind power.

1.1) Use the PSO-BP neural network algorithm to predict wind power deterministically, and obtain the deterministic prediction information of wind power.

As shown in Figure 3, it is a comparison chart between the predicted value of wind power and the actual value. The training sample adopted by the present invention is the measured data of the wind farm in the past year. As can be seen from Fig. 3, the average relative error MAE between the wind power deterministic prediction value and the actual value is 14.19%, and the root mean square error is 19.06 %. It can be seen that most of the relative errors of the wind power obtained by using the PSO-BP neural network algorithm fall within the range of -20% to +20%.

1.2) Using the kernel density estimation method in non-parametric estimation, the interval prediction of the wind power of the energy base system is carried out, and the interval prediction information of the wind power is obtained.

By analyzing the error between the output power value of the wind farm and its real value in Figure 3, it can be known that in the case of a large difference in wind power value, the fluctuation of wind power prediction error is also large, so we can use the predicted value The size of the wind power prediction error is divided into multiple power intervals, and the wind power prediction error distribution models of different prediction power intervals are established independently in turn.

The training data of the model uses the predicted value and actual value of wind power from June 1, 2015 to May 29, 2016, a total of 8736 wind power data samples, and the test data of the model uses May 30, 2016 to May 31 There are 48 wind power data samples in total.

As shown in Fig. 4 and Fig. 5, they are the probability density function of the prediction error fitting of a certain power interval and the 90% probability interval of wind power on May 30-31. From the figure, we can see that the accuracy of wind power prediction is not convincing from time to time, especially when the wind power output is relatively low, the relative error of wind power prediction is very large, and in the DC transmission system of the energy base, this kind of wind power The extreme situation of output is not uncommon, and this kind of situation is a great threat to the system. If we only use the pure wind power point forecast value to make the day-ahead plan, it will not be able to meet the needs of the system operation. It can be seen that the probability interval covers most of the wind power change information, and the interval prediction method can be relatively accurate in describing the uncertainty of the wind power point prediction value.

1.3) Use the scenario analysis method to model the uncertainty of wind power, and use the daily variation dimension scenario optimization method based on the mean value clustering method and the hour dimension scenario optimization method based on the tabu search method to optimize the wind power sequence scenario.

As shown in Figure 6(a)~(c), it is the basic idea of generating a two-dimensional optimization method that reflects the randomness and fluctuation of wind power power. Among them, Figure 6(a) represents a schematic diagram of the historical daily wind power sequence, the abscissa shows the time period of the wind power, the vertical axis shows the number of days it is in, and the circles represent the wind power value of the time period on that day. The first thing to do is to optimize the wind power sequence in the dimension of diurnal variation. In the schematic diagram, the optimization is carried out from the direction of the ordinate to generate representative scenarios that reflect the statistical characteristics of wind power in each period. Select a certain time period, such as the time period in the dashed box in Figure 6(a), and select several scenes that can represent this time period, as shown in Figure 6(b), three representative scenes are generated for each time period. The next thing to do is to optimize the wind power sequence in the dimension of hourly variation. In the schematic diagram, the optimization is carried out from the direction of the abscissa. Using the representative scenarios of each time period formed in the previous step, select the appropriate scenario, and place each The scenes selected by the period are connected to form multiple sequences from period 1 to period T, and the final wind power sequence scene is obtained. The final result is shown in Figure 6(c).

As shown in Figures 7 to 10, five representative scenarios are finally generated through the scene optimization algorithm of daily changing dimensions. Figure 7 and Figure 8 are the wind power prediction errors and the corresponding scene probabilities of the five optimal representative scenarios generated by the diurnal variation dimension, respectively. Figure 9 and Figure 10 are comparison diagrams of the average value and variance of the scenes in each time period generated by the daily change dimension optimization. The statistical characteristics of the representative scene optimized by using the daily change dimension scene are not very different from the original scene. To some extent, we can think that the optimized representative scene can reflect the statistical characteristics of the original scene.

As shown in Figures 11-13, based on the five optimal representative scenarios in each time period generated based on the diurnal variation optimization algorithm dimension scenarios, 50, 120 and 300 wind power sequence scenarios are generated respectively. Figure 11 is the trend change diagram of the fitness function. It can be seen from the figure that with the increase of the number of iterations, the value of the fitness function continues to increase, and finally tends to be stable, that is, the optimal wind power prediction error sequence has been found. Fig. 12 shows the probability values corresponding to each scene when 120 scene sequences are generated, and Fig. 13 shows that 20 scenes in the generated 120 scene sequences are selected and displayed.

2) Using the above various wind power forecast information, establish a day-ahead power generation planning model of the energy base system including DC regulation means.

3) Set multiple calculation examples to solve the day-ahead power generation plan model of the energy base system, and determine the optimal day-ahead power generation plan of the energy base system according to the comparison results of operating indicators and key indicators.

In order to verify the validity of the proposed method of the present invention, the design calculation example is as follows:

Calculation example 1: The deterministic prediction value of wind power utilization, and the reserve capacity is zero;

Calculation example 2: The deterministic prediction value of wind power utilization, the reserve capacity is 5% of the DC power in the corresponding period;

Calculation example 3: Wind power utilization is 80% of the deterministic forecast value, and the reserve capacity is zero;

Calculation example 4: The predicted value of wind power utilization interval, the reserve capacity ratio is zero;

Calculation example 5: Predicted value of wind power utilization scenario, the reserve capacity ratio is zero.

In each case scenario, the comparative indicators of the energy base system include:

①Operating cost of conventional unit: After determining the unit combination decision in each example scenario, it includes the generation cost of the thermal power unit in the dispatch period and the unit start-up cost in the unit combination decision.

②System cut-off wind power: In the calculation period, under the condition of actual wind turbine output, the system cut-off wind volume caused by the unit combination decision of each example scenario based on the wind power prediction value.

③The number of time periods for the system to generate wind curtailment: within the calculation period, under the actual output of wind turbines, the number of time periods for the system to curtail wind caused by the unit combination decision-making results.

④The number of time periods for the system to generate DC regulation: the number of time periods during which DC participates in the consumption of wind power to generate regulation within the calculation cycle.

⑤ Total operating cost of the system: the total operating cost of the system including the operating cost of conventional units, wind curtailment cost and DC regulation cost.

Table 1 shows different calculation scenarios. After obtaining the unit combination decision of each calculation scenario, based on the real output of wind power, the comparison of different operation indicators. Calculation example 4 adopts the probability interval of wind power prediction, and the amount of wind curtailment is less than that of the previous three scenarios, and at the same time, the number of DC regulation periods increases. In Calculation Example 5, the wind power scenario prediction information is used, and the operating cost of the system is significantly higher than that of Calculations 1, 2 and 4. The reason is that the wind power scenario contains some extreme wind power output conditions, which are quite different from the real wind power output conditions, resulting in The operating cost of the system is increased, but the amount of abandoned air is the least among all calculation examples, and the degree of utilization of direct current is also the highest. In the utilization of wind power forecasting information, interval forecasting and scenario analysis methods have different contributions to system operating costs, wind curtailment and DC utilization, but compared with conventional wind power deterministic forecasting information, various performances have been improved .

Table 1 Comparison of running indicators of each example

As shown in Figure 14, it is the comparison between the planned operation value before the DC adjustment and the adjusted planned operation value on a certain day. It can be seen from the figure that there are peaks in the wind power in period 14 and period 15, and the DC plan value before adjustment cannot adapt to this change at this time, and the conventional thermal power unit cannot be adjusted to an appropriate power value due to the limitation of its own operating factors. . The adjusted DC planning value can well meet the fluctuation of this kind of wind power and avoid a large amount of wind curtailment. However, when the wind power trough occurs from time period 8 to time period 11, the corresponding DC power plan value is also lowered accordingly, so as to avoid a large power shortage in the sending end system because the output of thermal power has reached the upper limit. From time period 18 to time period 20, although the power of wind power also has a trough, because the planned operating power of DC is low in this period, that is, sufficient adjustment margin has been provided for the thermal power unit, so the DC has not been adjusted in this period. , it can be seen that the model not only adopts DC regulation means to regulate when thermal power cannot balance the fluctuation of wind power alone, but also avoids frequent regulation of DC in unnecessary situations.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: it can still be Modifications are made to the technical solutions described in the foregoing embodiments, or equivalent replacements are made to some of the technical features, and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims (10)

1. a kind of Energy Base system generation schedule formulating method a few days ago, it is characterised in that include the following steps：

1) according to the predictive information of the wind power obtained in advance, Energy Base system generation schedule Optimized model a few days ago is established；

2) it includes to apply example using different type wind power prediction information to formulate several, a few days ago as Energy Base system The input variable of generating plan model, according to the operating index comparing result of obtained Energy Base system and the operation stressed Index determines the optimal generation schedule a few days ago of Energy Base system.

2. a kind of Energy Base system as described in claim 1 generation schedule formulating method a few days ago, it is characterised in that：The step It is rapid 1) in, generation schedule Optimized model includes object function and constraints, the target letter to the Energy Base system a few days ago Number includes the object function based on point prediction information and the object function based on scene prediction information；The constraints includes work( Rate Constraints of Equilibrium, spare capacity constraint, unit output constraint, unit ramping rate constraints, unit minimum available machine time constraint, machine The minimum downtime of group constrains and abandons air quantity constraint.

3. a kind of Energy Base system as claimed in claim 2 generation schedule formulating method a few days ago, it is characterised in that：The base It is in the object function of point prediction information：

In formula,For the cost of electricity-generating of fired power generating unit,For the start-up cost of fired power generating unit,For stopping for fired power generating unit Machine cost,For the adjustment cost of direct current,To abandon eolian, footmark i represents i-th fired power generating unit, and footmark t represents The t periods.

4. a kind of Energy Base system as claimed in claim 2 generation schedule formulating method a few days ago, it is characterised in that：The base It is in the object function of scene prediction information：

In formula,For the cost of electricity-generating of fired power generating unit,For the start-up cost of fired power generating unit,For stopping for fired power generating unit Machine cost,For the adjustment cost of direct current,To abandon eolian, footmark i represents i-th fired power generating unit, and footmark t represents T periods, footmark j represent jth scene, and T is period sum, and N is the total number of units of fired power generating unit, and S is scene sum, p_jFor scene j's Probability.

5. a kind of Energy Base system as claimed in claim 2 generation schedule formulating method a few days ago, it is characterised in that：It is each described The calculation formula of constraints is respectively：

1. power-balance constraint：

In formula：p_itIt is output of i-th unit in the t periods,For the wind power prediction value of t periods,For due to Power-balance constraint and generate abandon wind-powered electricity generation amount,WithFor the planned value and adjusted value of t period dc powers；

2. spare capacity constrains：

In formula：up_itAnd dn_itUpper spinning reserve capacities and lower spinning reserve capacity of the respectively unit i in the t periods,WithThe respectively upper and lower bound in wind power prediction section, r_resupAnd r_resdnSpare capacity ratio is lowered for upper reconciliation, r_uiAnd r_diThe creep speed up and down of respectively generator i limits, p_imaxAnd p_iminFor the minimum and maximum output of generator i, t_r For the spare capacity actuation time of system requirements, α is interval prediction reserve factor, when Optimized model is pre- using wind power section α=1 when measurement information, otherwise α=0；z_itState variable for unit i in the t periods, value are 0 or 1, z_itUnit i is indicated when being 1 It is switched on operation in the t periods；z_itIndicate that unit i is shut down in the t periods when being 0；

3. unit output constrains：

4. unit ramping rate constraints：

In formula, p_itIt is outputs of the unit i in the t periods；p_i(t-1)It is outputs of the unit i in the t-1 periods；△ t are time intervals；

5. unit minimum available machine time constraint：

In formula：UT_iWithWhen being the minimum booting operation of unit i respectively Between and initial time period be already powered on time of operation, Z_i0It is operating statuses of the unit i in the initial period, Z_i0=1 indicates at this time Unit is open state, Z_i0=0 indicates that unit is off-mode；

6. minimum downtime constraint：

In formula：DT_iWithWhen being the minimum shutdown of unit i respectively Between and section has been shut down at the beginning time；

7. abandoning air quantity constraint：

In formula：To abandon wind-powered electricity generation amount due to what power-balance constraint generated,It is insufficient due to lowering spare capacity And what is generated abandons wind-powered electricity generation amount.

6. a kind of Energy Base system as described in claim 1 generation schedule formulating method a few days ago, it is characterised in that：The step It is rapid 1) in, the predictive information of the wind power includes the deterministic forecast of wind power, interval prediction and scene prediction information.

7. a kind of Energy Base system as claimed in claim 6 generation schedule formulating method a few days ago, it is characterised in that：The wind The computational methods of the scene prediction information of electrical power, include the following steps：

First according to the historical data of the real output of wind power plant and prediction power, the prediction error of wind power is obtained Historical data, and generate initial wind-powered electricity generation error sequence scene；

Secondly, the diurnal variation dimension scene optimization method based on means clustering method is the period to the initial wind-powered electricity generation using hour Error sequence scene is clustered, and the representative scene set that can reflect this day day part wind power statistical property is generated；

Finally, the hour based on taboo search method changes dimension scene optimization method, utilizes the representative scene of this day day part Set, the scene chosen respectively in day part are attached, and wind power prediction error sequence are formed, after successive ignition Close sequence of scenes is rejected, the scene prediction value sequence of final wind power is obtained.

8. a kind of Energy Base system as described in claim 1 generation schedule formulating method a few days ago, it is characterised in that：The step It is rapid 2) in, the method for obtaining optimal generation schedule a few days ago includes the following steps：

2.1) it includes to apply example using different type wind power prediction information to formulate several；

2.2) it using each application example as input variable, is input in generation schedule Optimized model a few days ago, each application is calculated The operating index result of example；

2.3) analysis is compared to obtained different operating index results, the operation actually stressed according to Energy Base system Index determines the optimal optimal generation schedule a few days ago using example as Energy Base system.

9. generation schedule formulates system, feature to a kind of Energy Base system suitable for method as described in claim 1 a few days ago It is：Module and seismic responses calculated module are built including Optimized model；

The Optimized model builds module, for the predictive information according to the wind power obtained in advance, establishes Energy Base system System generation schedule Optimized model a few days ago；

The seismic responses calculated module, for formulating several application calculations comprising using different type wind power prediction information Example refers to as the input variable of Energy Base system generating plan model a few days ago according to the operation of obtained Energy Base system Mark comparing result and the operating index stressed, determine the optimal generation schedule a few days ago of Energy Base system.

10. generation schedule formulates system to a kind of Energy Base system as claimed in claim 9 a few days ago, it is characterised in that：It is described Optimized model structure module includes predictive information acquisition module, object function structure module and constraints structure module；It is described Predictive information acquisition module is used to calculate deterministic forecast, interval prediction and the scene prediction information of wind power；The target Function builds module and is used to build object function according to the different type of wind power prediction information；The constraints builds mould Block is used to establish the related constraint function of object function；

The seismic responses calculated module includes that example formulates module, seismic responses calculated module and operating index comparison mould Block；It includes to apply example using different type wind power prediction information that the example, which formulates module and is used to formulate,；It is described excellent Change model computation module to be used to that operating index result to be calculated according to different application example；The operating index comparing result is used In being compared analysis to obtained different operating index results, optimal application example is obtained as optimal power generation meter a few days ago It draws.