Active power distribution network robust optimization scheduling method and device

Abstract

The invention provides an active power distribution network robust optimization scheduling method and device, and belongs to the technical field of distribution network collaborative optimization scheduling. According to the method, the photovoltaic output data can be predicted through the SARIMA model, the photovoltaic output distribution range is determined as the uncertainty set, and the uncertainty degree of the uncertainty set is represented through the confidence coefficient of the SARIMA model. Furthermore, robust optimization solution is carried out based on the improved photovoltaic output uncertainty set, a day-ahead scheduling result is obtained, and scheduling control is carried out. After improvement, the probability that photovoltaic output appears in the distribution range of the uncertain set is large, and the probability that photovoltaic output appears outside the distribution range of the uncertain set is small, so that the disadvantage of blind optimization in a box type interval during robust optimization is avoided, and the conservative property of a robust optimization scheme is reduced. Meanwhile, the range of the uncertain set better conforms to an actual photovoltaic output scene, the range of the uncertain set is narrowed, and the solution efficiency of robust optimization is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of coordinated optimization and dispatching of distribution networks, and in particular to a method and device for robust optimization and dispatching of active distribution networks. Background Art

[0002] The penetration rate of renewable energy such as photovoltaics in the power system will continue to increase. However, the large-scale development and high-proportion access of distributed photovoltaics have exposed the distribution network to the risk of random power fluctuations, causing problems such as power reverse transmission and voltage over-limit on the user side. Traditional distribution networks need to shift towards source-load interaction and flexible consumption, and their dispatching and operation methods also need to promote the diversification of energy replacement. Therefore, how to effectively predict the output uncertainty of renewable energy in the dispatching of distribution networks with a high proportion of distributed photovoltaics and improve the consumption capacity under the premise of ensuring the safe and economic operation of the distribution network has become one of the problems that need to be solved urgently.

[0003] The high proportion of distributed photovoltaic access and the mining of diversified demand response resources (DR) have caused a profound change in the "source-load" system. The distribution network is accelerating its transformation towards an active distribution network system, and it is urgently needed to provide a more reliable scheduling execution solution to improve the absorption and processing capacity of distributed photovoltaics. At present, the approaches to solving the uncertain optimization problems caused by distributed photovoltaics are mainly divided into two categories: stochastic programming and robust optimization. Robust optimization avoids the dimensionality curse of high-dimensional solutions in stochastic programming and uses uncertain intervals to describe photovoltaic output, so there is no need to build a deviation model, which has been increasingly favored.

[0004] The uncertainty set of traditional robust optimization is usually a box-type interval. The robust optimization is used to construct a box-type interval of photovoltaic output containing its historical data, and the optimization problem is solved on the boundary of the interval (i.e., the "worst" case), and the results obtained are relatively conservative. In reality, the probability that the uncertain variable photovoltaic output takes the limit value of the box-type interval boundary is low. Therefore, based on the uncertainty set of the traditional box-type interval, the robust optimization solution obtained by the active distribution network scheduling model is overly conservative and insufficiently practical, and the efficiency of robust optimization is low. Summary of the invention

[0005] The present invention provides a method and device for robust optimization scheduling of an active distribution network, so as to solve the problems of excessive conservatism and low robust optimization efficiency of the existing method for robust optimization scheduling of an active distribution network.

[0006] In a first aspect, the present invention provides a method for robust optimization scheduling of an active distribution network, comprising: obtaining historical photovoltaic output data of distributed photovoltaics in an active distribution network; based on the historical photovoltaic output data, predicting the photovoltaic output data with a preset confidence through a SARIMA model; determining the photovoltaic output distribution range based on the predicted photovoltaic output data, and using it as an uncertainty set of photovoltaic output, wherein the preset confidence represents the degree of uncertainty of the uncertainty set; based on the uncertainty set of photovoltaic output, with the goal of minimizing the scheduling cost of adjustable resources of the active distribution network, robustly optimizing and solving the day-ahead scheduling model of the active distribution network to obtain a day-ahead scheduling result of the active distribution network; wherein the day-ahead scheduling result includes the planned output power of the adjustable resources; and scheduling and controlling the active distribution network based on the day-ahead scheduling result.

[0007] In a second aspect, the present invention provides a robust optimization scheduling device for an active distribution network, comprising: an acquisition module for acquiring historical photovoltaic output data of distributed photovoltaics in the active distribution network; a prediction module for predicting photovoltaic output data based on the historical photovoltaic output data through a SARIMA model with a preset confidence; an uncertainty set acquisition module for determining the photovoltaic output distribution range based on the predicted photovoltaic output data, and using it as an uncertainty set of photovoltaic output, wherein the preset confidence represents the degree of uncertainty of the uncertainty set; a solution module for performing robust optimization solution on the day-ahead scheduling model of the active distribution network based on the uncertainty set of the photovoltaic output and taking the scheduling cost of the adjustable resources of the active distribution network as the minimum, to obtain the day-ahead scheduling result of the active distribution network; wherein the day-ahead scheduling result includes the planned output power of the adjustable resources; and a scheduling module for scheduling and controlling the active distribution network based on the day-ahead scheduling result.

[0008] The present invention provides a method and device for robust optimization scheduling of an active distribution network, which predicts photovoltaic output data through a SARIMA model, determines the distribution range of photovoltaic output as an uncertainty set, and represents the degree of uncertainty of the uncertainty set through the confidence of the SARIMA model. Furthermore, based on the uncertainty set of the improved photovoltaic output, robust optimization is solved, the day-ahead scheduling result is obtained, and scheduling control is performed. Due to the improvement, the probability of photovoltaic output appearing within the distribution range of the uncertainty set is greater, and the probability of appearing outside the distribution range of the uncertainty set is smaller, which avoids blindly searching for advantages and disadvantages in the box interval during robust optimization, thereby reducing the conservatism of the robust optimization scheme. At the same time, the range of the uncertainty set is more in line with the actual photovoltaic output scenario, the range of the uncertainty set is reduced, and the efficiency of robust optimization solution is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0010] Figure 1 is an application scenario diagram of the active distribution network robust optimization scheduling method provided by an embodiment of the present invention;

[0011] Figure 2 It is a structural schematic diagram of a robust optimization box interval in the prior art;

[0012] Figure 3 is a flowchart of an implementation method of a robust optimization scheduling method for an active distribution network provided by an embodiment of the present invention;

[0013] Figure 4 is a schematic diagram of the structure of an uncertain set provided by an embodiment of the present invention;

[0014] Figure 5 is a nonlinear affine robust optimization flow chart provided by an embodiment of the present invention;

[0015] Figure 6 It is a two-stage uncertainty scheduling framework diagram of an active distribution network provided by an embodiment of the present invention;

[0016] Figure 7 is a real-time affine MPC scheduling optimization flow chart provided by an embodiment of the present invention;

[0017] Figure 8 It is a structural diagram of an active distribution network robust optimization scheduling device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.

[0019] In order to make the purpose, technical solutions and advantages of the present invention more clear, specific embodiments will be described below in conjunction with the accompanying drawings.

[0020] Figure 1 This is an application scenario diagram of the active distribution network robust optimization scheduling method provided by the embodiment of the present invention. Figure 1As shown, the active distribution network usually includes a source side and a load side. For example, the source side may include distributed photovoltaic and power storage systems; the load side may include fixed loads and demand-side response resources.

[0021] The dispatch model of active distribution network is used to coordinate and dispatch various resources. The dispatch of distribution network with high proportion of distributed photovoltaic is affected by the uncertainty of photovoltaic output. At present, the approaches to solve the uncertain optimization problems caused by distributed photovoltaic are mainly divided into two categories: stochastic programming and robust optimization.

[0022] The stochastic programming method constructs a random scenario tree based on the future prediction of uncertain variables, and obtains the optimal solution by optimizing all possible paths of the scenario tree. The existing technology constructs a probability distribution prediction model of wind power and photovoltaic space to form a scenario tree in the scheduling time domain, and obtains the optimal solution through cyclic iterative optimization. However, with the increase of distributed photovoltaic nodes and the improvement of the dispersion of photovoltaic output, the number of future prediction scenarios has increased exponentially, resulting in long solution time and low efficiency. Although the improved chance constrained stochastic optimization method avoids the problem of high-dimensional solution, the uncertain variables must satisfy a certain probability distribution to specify the confidence level. The actual historical output data of distributed photovoltaics often lacks correlation and has a large deviation from the constructed distribution model. Robust optimization can better avoid the defects of stochastic programming methods.

[0023] Figure 2 It is a structural diagram of the robust optimization box interval in the prior art. Figure 2 , the horizontal axis represents time, the unit can be hours, the range can be one day; the vertical axis represents the output of distributed photovoltaics. The box interval takes the fixed photovoltaic output range as the uncertainty range at each moment, and the final uncertainty set is Figure 2 In the middle, it is in the shape of a rectangular box. Generally, the photovoltaic output box interval is constructed to include its historical data. For example, according to the historical data of the original photovoltaic output, the box interval boundary for optimization during uncertain optimization is formed. The uncertain variable set covers all output conditions as much as possible to form a quadrilateral interval boundary. At the same time, the interval includes a large amount of blank space, which increases the blindness of the optimization path. Robust optimization solves the optimization problem on the interval boundary (i.e., the "worst" case), and the results obtained are relatively conservative. In reality, the probability of uncertain variables obtaining the limit value of the interval boundary is low, so the optimization solution obtained by traditional box-type robust optimization is generally overly conservative and insufficiently practical.

[0024] At present, in order to overcome the conservatism and blindness of robust optimization in the prediction interval of uncertain variables, existing literature has explored two aspects: interval modeling and dual solution. However, existing research focuses on the spatiotemporal characteristics of distributed photovoltaic output intervals, and lacks seasonal and long-term processing of historical data output. The interval boundaries obtained are relatively conservative and do not match the actual photovoltaic output in the area.

[0025] The embodiment of the present invention constructs an uncertainty set based on the prediction results of the SARIMA model, avoiding the blind search for advantages and disadvantages in the box interval during robust optimization, thereby reducing the conservatism of the robust optimization scheme. The range of the uncertainty set is more consistent with the actual photovoltaic output scenario.

[0026] Figure 3 The implementation flow chart of the active distribution network robust optimization scheduling method provided by the embodiment of the present invention is detailed as follows:

[0027] Step 301, obtaining historical photovoltaic output data of distributed photovoltaics in an active distribution network.

[0028] In some embodiments, steps 301-305 may be performed in the day-ahead stage. For example, they may be performed before the dispatching day. Accordingly, the historical photovoltaic output data of the distributed photovoltaics before the dispatching day in the active distribution network may be obtained. It should be noted that the dispatching day is the date on which the dispatching of the active distribution network is actually performed. It is not limited to a full day here. For the convenience of explanation, the subsequent steps take the dispatching day as the day and the date before the dispatching day as the day-ahead.

[0029] In some embodiments, the distributed photovoltaic includes a plurality of photovoltaic nodes. The historical photovoltaic output data may be photovoltaic output data of each photovoltaic node.

[0030] In some embodiments, the historical photovoltaic output data includes the historical photovoltaic output power of the photovoltaic node.

[0031] Step 302: Based on the historical photovoltaic output data, the photovoltaic output data is predicted by using the SARIMA model with a preset confidence level to obtain photovoltaic output data.

[0032] In this scheme, the SARIMA model is used to predict photovoltaic output data. The following first introduces the SARIMA model.

[0033] SARIMA (Seasonal Autoregressive Movingaverage, SARIMA for short) forecasting method is improved by adding seasonal factors to the ARIMA forecasting model. SARIMA includes the following elements: non-seasonal autoregressive coefficient, difference coefficient, non-seasonal moving average order, seasonal autoregressive coefficient, seasonal difference number, seasonal moving average coefficient

[0034] AR(p): Non-seasonal autoregressive coefficient. Used to simulate the linear relationship between the current observation and the past observation. It uses a linear combination of p lagged observations to predict the current observation x. t , which is expressed as follows:

[0035] x t=Φ 0 +Φ 1 x t-1 +…+Φ p x t-p +ε t

[0036] d: difference coefficient. It is used to transform non-stationary time series data into stationary series through multi-order differences.

[0037] MA(q): non-seasonal moving average order. It is used to simulate the linear relationship between the current observation error and the past observation error. It uses the linear combination of q lagged errors to predict the current observation value, expressed as:

[0038] x t =θ 0 +θ 1 ε t-1 +θ 2 ε t-2 +…+θ q ε t-q +ε t

[0039] Φ P : Seasonal autoregression coefficient, P is the order, which is used to quantify the impact of past seasonal cycles in the time series on the current value.

[0040] x t =α 0 +α 1 ε t-s +…+α P ε t-Ps +ε t

[0041] D is the seasonal differencing order. It makes seasonal data stationary by differencing the data for a specific seasonal period length.

[0042] Θ Q is the seasonal moving average coefficient, and Q is its order. It is used to describe how the error term in the past seasonal cycle affects the observed value of the current cycle, expressed as follows

[0043] x t =β 0 +β 1 ε t-s +β 2 ε t-2s +…+β q ε t-Qs +ε t

[0044] In summary, the mathematical model of SARIMA is as follows:

[0045] φ(p) (B)Φ (P) (B s )(1-B) d (1-B s ) D x t =θ (q) (B)Θ (Q) (B s )ε t

[0046] Among them, θ 1 ,…,θ q is the moving average coefficient; B is the difference operator, B n x t =x t-n ; s is the seasonal cycle length, which represents the period of the recurring seasonal pattern in the time series data; ε t is the error value.

[0047] The generalized SARIMA seasonality refers not only to the law of historical data changing with the seasons, but also to any type of periodic change law in the observation time domain. s is the change period of the observation data. Taking the photovoltaic output data on the dispatch day as an example, s = 24.

[0048] It should be noted that based on the historical photovoltaic output data, the photovoltaic output data predicted by the SARIMA model can include the probability of a photovoltaic output at a certain moment. The confidence level is the degree of certainty about the probability of photovoltaic output. The above-mentioned preset confidence level can be determined based on the target uncertainty of the robust optimization uncertainty set. The relationship between confidence level and uncertainty is explained in the following steps.

[0049] The embodiment of the present invention constructs an uncertainty set by predicting photovoltaic output data through the SARIMA model, taking into account the long-term seasonal fluctuation law of distributed photovoltaic historical data, and the obtained data is more consistent with the actual photovoltaic output.

[0050] Step 303: Determine the distribution range of photovoltaic output based on the predicted photovoltaic output data and use it as an uncertainty set of photovoltaic output, wherein the preset confidence level represents the degree of uncertainty of the uncertainty set.

[0051] In some embodiments, for a certain period of time, the distribution range of photovoltaic output may be from the minimum photovoltaic output to the maximum photovoltaic output in the period of time. Alternatively, for a certain moment, the distribution range of photovoltaic output may be from the minimum photovoltaic output to the maximum photovoltaic output at that moment.

[0052] In some embodiments, the photovoltaic output distribution range represents the range in which the photovoltaic output in the predicted data may appear under a preset confidence level. The probability of the actual photovoltaic output appearing within this range is high, while the probability of appearing outside this range is low. Therefore, the photovoltaic output distribution range can be used as an uncertainty set of photovoltaic output for robust optimization in subsequent steps.

[0053] Regarding the relationship between confidence and uncertainty, it should be noted that the greater the preset confidence used in the prediction, the larger the photovoltaic output distribution range of the obtained photovoltaic output data, and the greater the uncertainty of the uncertainty set thus constructed.

[0054] Figure 4 is a schematic diagram of the structure of an uncertain set provided by an embodiment of the present invention; Figure 4 , the horizontal axis represents time; the vertical axis PV represents photovoltaic output; each solid line curve and its nearby points in the figure represent an uncertainty set under a certain confidence level, that is, an uncertainty set with different degrees of uncertainty. Indicates different degrees of uncertainty.

[0055] The following example illustrates how to obtain an uncertainty set from photovoltaic output data.

[0056] In a possible implementation, the photovoltaic output distribution range is determined based on the predicted photovoltaic output data, and the photovoltaic output uncertainty set includes:

[0057] Step 3031: for any prediction period, determine the extreme power point of the photovoltaic output data within the prediction period.

[0058] Exemplarily, the extreme power points of a certain period include a maximum value and a minimum value. It should be noted that using a period as a processing unit can avoid the volatility of the forecast data at a single moment, and the final envelope is smoother and more in line with the actual photovoltaic output.

[0059] Step 3032: based on the extreme power points of the photovoltaic output data in each time period, obtain the envelope of the photovoltaic output data, wherein the envelope includes a maximum envelope and a minimum envelope.

[0060] In some embodiments, the maximum values ​​of each time period are connected to obtain a maximum value envelope; the minimum values ​​of each time period are connected to obtain a minimum value envelope.

[0061] Step 3033: determine the range between the maximum envelope line and the minimum envelope line as the photovoltaic output distribution range.

[0062] Step 3034, taking the photovoltaic output distribution range as the photovoltaic output uncertainty set.

[0063] The following embodiments illustrate the specific form of the uncertain set of photovoltaic output.

[0064] In a possible implementation, the expected curve of the predicted photovoltaic output data and the predicted photovoltaic output fluctuation range constitute an uncertainty set of photovoltaic output, wherein the confidence of the prediction model multiplied by the fluctuation value of the photovoltaic output obtains the predicted photovoltaic output fluctuation range.

[0065] The following is another embodiment to illustrate the specific form of the uncertainty set of photovoltaic output. Distributed photovoltaic includes multiple photovoltaic nodes, and the predicted data includes multiple time periods.

[0066] In a possible implementation, the uncertainty set of photovoltaic output is:

[0067]

[0068] Among them, Ω pv represents the uncertainty set of PV output; represents the expected value of the photovoltaic output at node j during period t; represents the fluctuation value of photovoltaic output at node j during period t; ρ j,t Represents the confidence of the SARIMA model at node j in period t; is the uncertainty adjustment coefficient of distributed photovoltaics, which represents the upper limit of the sum of the confidence levels of all nodes of distributed photovoltaics.

[0069] It should be noted that the expected value of the photovoltaic output at node j in period t is the average value of the photovoltaic output at node j in period t. It can be a covariance vector, which can be obtained by analyzing the mathematical model of SARIMA. The confidence of each node of distributed photovoltaic is not necessarily the same, and the confidence can be set independently. Indicates the uncertainty degree of the distributed photovoltaic uncertainty set. The larger the value, the higher the envelope of the photovoltaic output extreme value, the larger the internal output range, and the greater the uncertainty of optimization: Uncertainty adjustment coefficient The larger the value, the greater the uncertainty of the optimization. Since distributed photovoltaics include multiple nodes, the confidence of each node is added to obtain the uncertainty of the distributed photovoltaics as a whole. In the above formula, the sum of the confidence of each node should be less than or equal to the uncertainty of the uncertainty set.

[0070] The expected curve and confidence interval of the uncertainty set in the embodiment of the present invention are highly consistent with the photovoltaic output. Different uncertainty adjustment coefficients can be achieved by adjusting the confidence of the SARIMA model. Then the envelope range of photovoltaic output prediction is adjusted to form uncertainty optimization search intervals with different uncertainties.

[0071] In some embodiments, the uncertainty set of photovoltaic output may also be:

[0072]

[0073] Among them, z j,t A two-dimensional variable factor indicating whether photovoltaic power is connected at node j in period t.

[0074] The above steps lead to the uncertainty set of PV output. The following steps illustrate the robust optimization solution based on the uncertainty set.

[0075] Step 304, based on the uncertain set of photovoltaic output, with the goal of minimizing the dispatching cost of the adjustable resources of the active distribution network, the day-ahead dispatching model of the active distribution network is robustly optimized and solved to obtain the day-ahead dispatching result of the active distribution network; wherein the day-ahead dispatching result includes the planned output power of the adjustable resources.

[0076] The following first explains the active distribution network dispatch and dispatch model. The distribution network dispatch center can coordinate the upper regional power grid, local dispatchable resources ESS (electric energy storage system), DR, etc., to smooth the impact of distributed photovoltaic prediction errors on system power balance at multiple time scales such as day-ahead and real-time.

[0077] For example, the active distribution network day-ahead dispatch model can construct an optimization objective function F based on the distribution network power purchase cost, ESS cycle life depreciation cost, DR response cost, and network loss cost. The objective function can be in the following formula form:

[0078]

[0079] Where T is the scheduling period; c loss represents the network loss cost coefficient; Indicates the electricity price purchased from the main grid; c pv represents the DR response cost coefficient; Ω L represents the branch set; Ω tr Represents the distribution transformer node set; Ω DR Represents the set of photovoltaic access nodes; I ij,t is the current in the line between nodes i and j during period t; is the transaction power on node j during period t; C ess It is the cycle life depreciation cost of ESS energy storage.

[0080] For example, the cycle life loss of ESS refers to the loss of mechanical life during the cyclic charging and discharging process within the dispatching day, which is generally determined by its rated capacity and rated charging and discharging power. ess Calculated from the annual life cycle cost:

[0081]

[0082] Among them, Ω ess is the set of access nodes of ESS; represents the rated capacity of ESS on node j; represents the rated charging and discharging power of ESS on node j; are the annual investment cost coefficients of ESS unit capacity and unit charging and discharging power, respectively. The calculation formulas refer to the above formula; are the cost coefficient of the unit capacity of the ESS at the node and the cost coefficient of the unit power of the ESS at the node respectively; r is the discount compensation rate; is the ESS cycle life of the node.

[0083] The life cycle loss cost of ESS is C ess After linearization, we get:

[0084]

[0085] in, are the linear fitting coefficients; is the equivalent number of cycles of ESS based on the depth of discharge.

[0086] The above part introduces the cycle life loss cost C of ESS in the active distribution network day-ahead dispatch model. ess ,The constraints of the day-ahead scheduling model are described below.

[0087] For example, the constraints of the day-ahead dispatch model may include DR-type load constraints, second-order cone relaxation power balance constraints, node voltage and current constraints, ESS operation constraints, reactive power compensation equipment power constraints, and distribution network node gateway power constraints. Each constraint is described below one by one.

[0088] 1) DR type load constraints. Node users in the distribution network are aggregated into different types of DR according to their power consumption characteristics to participate in distribution network scheduling. The load on node j is aggregated into the following four categories according to its response mode: basic load, curtailable load, transferable load and shiftable load.

[0089] ① The basic load is the part of the user load that cannot change the power consumption plan, recorded as The system needs to fully meet its electricity demand and is a non-dispatchable resource.

[0090] ② Class I: The curtailable load is the user load on the node that can withstand a certain interruption or reduce power demand, denoted as Under the incentive of the compensation policy, it reduces part or all of the load according to the actual supply and demand. Its mathematical model is as follows:

[0091]

[0092] Where: γ t The two-dimensional variable is the reduction status of the load that can be reduced during period t. A value of 1 indicates that the load is reduced during this period, and vice versa. It is the reduction coefficient during period t, θ t = [0, 1]; is the maximum load that is planned to be reduced.

[0093] The compensation cost of reducing load to participate in dispatch is:

[0094]

[0095] Where: It is the compensation cost coefficient of unit load reduction.

[0096] ③ Category II: The transferable load is the user load on the node that can bear the power demand transfer period, recorded as Under the incentive of the compensation policy, the load is flexibly transferred to other periods within the dispatch cycle, but the total load remains constant:

[0097]

[0098] Where: β t The two-dimensional variable is the reduction status of the transferable load in period t, where a value of 1 means that it is transferred in this period, and vice versa; are the upper and lower limits of the transferable load in period t respectively.

[0099] The compensation cost of transferable load participating in dispatch is:

[0100]

[0101] Where: It is the compensation cost coefficient of unit load reduction.

[0102] In addition, the response period of the transferable load should be constrained to avoid a load being transferred to multiple periods and frequent start and stop, which can be expressed as follows:

[0103]

[0104] Where: It is the minimum continuous operating time period for transferring load.

[0105] ④ Class III: The shiftable load is similar to the transferable load, but the load needs to be shifted as a whole when the load power consumption period can be responded according to the plan, while maintaining the original power consumption time, recorded as The translation model is as follows:

[0106]

[0107] With two-dimensional variable α t Indicates the translation state of period t, and a value of 1 indicates translation starting from period t. The load model after translation is:

[0108]

[0109] The response compensation cost of the translatable load is:

[0110]

[0111] Where: is the compensation cost coefficient per unit of curtailable load; is the sum of the translational load powers.

[0112] In summary, the mathematical model of demand response load is:

[0113]

[0114] That is, demand response load = load that can be reduced + load that can be transferred + load that can be shifted.

[0115] The above describes the DR type load constraint, and the following describes the power balance constraint of the second-order cone relaxation.

[0116] 2) Power balance constraints of second-order cone relaxation. The power flow constraints of the distribution network include active power flow and reactive power flow. The mathematical model after second-order cone relaxation is as follows:

[0117]

[0118] Where: U(j) and V(j) are the node sets when the distribution network takes j as the terminal and the head node respectively; P ij,t , Q ij,t are respectively the active and reactive power of the branch between nodes ij during period t; P j,t , Q j,t are the active and reactive power injected into node j, respectively; are the charging and discharging power of the ESS connected to node j; are the active and reactive power of transaction, PV and load respectively; r ij 、x ij are the resistance and reactance of branch ij respectively; U i,t The voltage amplitude of node i at time t; It is the reactive power of the reactive compensation equipment.

[0119] 3) Node voltage and current constraints:

[0120]

[0121] Where: u j,max 、u j,min are the upper and lower limits of the voltage at node j; I ij,max It is the maximum current on the interconnection line between nodes i and j.

[0122] 4) ESS operation constraints. ESS has charging, discharging or dormant states during the scheduling cycle, and can only be in one of these states during a certain period of time. During operation, it is also subject to the upper and lower limits of charge, charge and discharge power, etc. The calculation formula is as follows:

[0123]

[0124] Where: are two-dimensional variables representing the charging and discharging states of the ESS, respectively, and a value of 1 indicates that it is in that state; They are the upper and lower limits of charging and discharging power respectively; Soc j,t is the charge capacity of the ESS on node j during period t, Soc j,t are their maximum and minimum capacity limits respectively; η ch,j , η dis,j They are charge and discharge efficiency respectively.

[0125] 5) Reactive power compensation equipment power constraint. The main function of reactive power compensation equipment is to maintain the stability of node voltage. In particular, the power reverse transmission in high-proportion photovoltaic distribution network systems increases the risk of voltage exceeding the limit at the user end. A continuous reactive power compensation device SVC is used, and its constrained model is as follows:

[0126]

[0127] Where: are the upper and lower limits of the SVC compensation power connected to node j.

[0128] 6) Power constraints at distribution network nodes. The power at the connection node between the distribution network and the main grid must be controlled within a certain range to avoid large fluctuations in the connection line power that disrupt the balance of the main grid.

[0129]

[0130] Where: P j,max , P j,min , Q j,nax , Q j,min are the upper and lower limits of the active power and reactive power at the node j respectively.

[0131] The above embodiments illustrate the objective function and constraints of the day-ahead dispatch model. Further, based on the uncertain set of photovoltaic output, with the goal of minimizing the dispatch cost of the adjustable resources of the active distribution network, the day-ahead dispatch model of the active distribution network is robustly optimized and solved to obtain the day-ahead dispatch result of the active distribution network. For example, the planned output power of the adjustable resources is obtained.

[0132] Step 305: Based on the day-ahead dispatch result, the active distribution network is dispatched and controlled.

[0133] For example, based on the day-ahead scheduling results, equipment start and stop strategies and DR response power consumption plans and power purchase and sales plans are formulated in the day-ahead stage.

[0134] The embodiment of the present invention predicts photovoltaic output data through the SARIMA model, determines the distribution range of photovoltaic output as an uncertainty set, and represents the degree of uncertainty of the uncertainty set through the confidence of the SARIMA model. Furthermore, based on the uncertainty set of the improved photovoltaic output, robust optimization is solved, the day-ahead scheduling result is obtained, and scheduling control is performed. Due to the improvement, the probability of photovoltaic output appearing within the distribution range of the uncertainty set is greater, and the probability of appearing outside the distribution range of the uncertainty set is smaller, which avoids blindly searching for advantages and disadvantages in the box interval during robust optimization, thereby reducing the conservatism of the robust optimization scheme. At the same time, the range of the uncertainty set is more in line with the actual photovoltaic output scenario, the range of the uncertainty set is reduced, and the efficiency of robust optimization solution is improved.

[0135] The embodiment of the present invention takes into account the long-term seasonal fluctuation law of distributed photovoltaic historical data and proposes a SARIMA interval prediction model to describe its output range. The optimization process of the embodiment of the present invention introduces an affine mechanism to characterize the uncertainty of distributed photovoltaic output interval prediction, establishes an affine mechanism between uncertain variables and controllable variables, reduces the disadvantage of blindly seeking the best in the interval, thereby reducing the conservatism of the optimization scheme and making it more in line with the actual photovoltaic output scenario.

[0136] After the confidence interval of the SARIMA prediction model is converted into an affine adjustable coefficient uncertainty for uncertainty optimization, the optimization interval is greatly reduced and the operation efficiency of the operation scheme is improved.

[0137] The above embodiments perform robust optimization solutions based on the uncertainty set of photovoltaic output, thereby reducing the conservatism of the robust optimization scheme. The following embodiments propose an affine adjustable robust optimization model based on SARIMA prediction confidence level, establish an affine mechanism between adjustable decision variables and random variables (e.g., photovoltaic output), further reduce the adjustment range of decision variables to avoid irregular iterations, and reduce the blindness of the optimization process.

[0138] In a possible implementation, based on the uncertain set of photovoltaic output, the day-ahead dispatch model of the active distribution network is robustly optimized to minimize the dispatch cost of the adjustable resources of the active distribution network. The day-ahead dispatch results of the active distribution network include:

[0139] Step 3041, based on the uncertainty set of photovoltaic output, the affine relationship between photovoltaic output and the adjustable decision variables of the day-ahead dispatch model of the active distribution network, an affine uncertainty set of the adjustable decision variables is obtained.

[0140] Exemplarily, the adjustable decision variables are used to describe the adjustable resources. There is a correlation between the adjustable resources and the photovoltaic output. For example, the greater the photovoltaic output, the more adjustable resources are required. Thus, an affine relationship between the photovoltaic output and the adjustable decision variables of the day-ahead dispatch model of the active distribution network can be established. The form of the affine relationship is described below.

[0141] In one possible implementation, the affine relationship between photovoltaic output and adjustable decision variables is:

[0142]

[0143] in, represents an adjustable decision variable; represents the initial value of the adjustable decision variable; f j (t) represents the preset affine proportional relationship; Indicates photovoltaic output.

[0144] Step 3042, based on the affine relationship between the photovoltaic output and the adjustable decision variables, and the day-ahead dispatch model of the active distribution network, a robust optimization dispatch model is obtained.

[0145] Exemplarily, the affine relationship between photovoltaic output and adjustable decision variables can be used to represent the adjustable decision variables of the day-ahead dispatch model of the active distribution network, thereby obtaining a robust optimization dispatch model.

[0146] Step 3043, based on the uncertainty set of photovoltaic output and the affine uncertainty set of adjustable decision variables, with the goal of minimizing the dispatching cost of adjustable resources of the active distribution network, the robust optimization dispatching model is dually solved to obtain the day-ahead dispatching result of the active distribution network.

[0147] Exemplarily, the affine uncertainty set of adjustable decision variables can also be obtained from the uncertainty set of photovoltaic output and the above affine relationship. Therefore, the uncertainty set of photovoltaic output and the affine uncertainty set of adjustable decision variables can constitute the robust optimization uncertainty set of the day-ahead dispatch model of the entire active distribution network.

[0148] The uncertainty of the uncertainty set of photovoltaic output is related to the confidence of SARIMA prediction. Therefore, the uncertainty of the uncertainty set of photovoltaic output and the uncertainty set of adjustable decision variables can be realized by adjusting the confidence of the SARIMA model.

[0149] The embodiment of the present invention establishes an affine mechanism between adjustable decision variables and photovoltaic output through an affine adjustable robust optimization model based on SARIMA prediction confidence level, further reduces the adjustment range of decision variables to avoid irregular iteration, and reduces the blindness of the optimization process.

[0150] Under this affine mechanism, the distribution network optimization problem is summarized as a robust optimization model in vector form:

[0151]

[0152] Where: and are the coefficients corresponding to the objective function; R, G, A c , B c , D c , C c , t, h c ,ω c are the variable coefficients corresponding to the constraints; x, y c They are the non-adjustable decision vector and the adjustable decision vector before and after the uncertain variables are realized, and their elements are as follows:

[0153]

[0154] The third constraint of the above-mentioned robust optimization model formula in vector form is a robust constraint formula containing uncertain variables. For the distribution network, the following formula must be satisfied to obtain an optimal solution within the robust interval of uncertain variables.

[0155]

[0156] Where: are the coefficients of the i-th row and j-th column respectively; x j ,y j are the j-th variable elements respectively.

[0157] The terms containing uncertain variables in the above vector form of robust optimization model formula are processed by scaling and relaxation. It can be solved directly. Therefore, the auxiliary variable is introduced Transform its terms into the following model:

[0158]

[0159] The duality theory is used to transform the above problem into a robust equivalence problem to solve its duality. After the above max problem is transformed into a min problem and summarized into the objective function, the uncertain optimization problem is transformed into a deterministic mixed integer optimization model and can be solved directly.

[0160] Figure 5 is a nonlinear affine robust optimization flow chart provided by an embodiment of the present invention. Figure 5 For example, first, a mathematical model for optimizing uncertainty in the distribution network is pre-built, such as a day-ahead dispatching model. Then, nonlinear affine is performed, including: using the SARMA random model to describe the uncertainty set of photovoltaic output, and establishing an affine relationship between photovoltaic output and adjustable decision variables; linearizing the nonlinear affine relationship, and constructing an adjustable model of random variables. Finally, based on the CCG dual solution, a solution result is obtained, such as a day-ahead dispatching result.

[0161] The above embodiment illustrates the use of affine robust optimization in the day-ahead stage. The following describes how to perform robust optimization in the intraday stage. First, the two-stage uncertainty dispatch framework of the active distribution network is described.

[0162] Figure 6 is a two-stage uncertainty scheduling framework diagram of an active distribution network provided by an embodiment of the present invention. Figure 6 ,The scheduling framework includes the day-ahead phase and the real-time ... can also be called the intraday phase.

[0163] In the day-ahead phase, equipment start-up and shutdown strategies, as well as DR response power consumption plans and power purchase and sales plans are formulated. In the real-time phase, ESS and DR are guided to coordinate and forecast loads to absorb distributed photovoltaic power as much as possible. For users participating in demand response, multiple response types are aggregated according to their power consumption characteristics to participate in power regulation at different time scales.

[0164] For example, in the day-ahead stage, historical photovoltaic data is obtained, and the output interval based on the confidence level, that is, the uncertainty set of photovoltaic output, is constructed based on the SARIMA time series analysis forecast. The day-ahead economic dispatch optimization solution is solved to obtain the action state and output of slow adjustable resources, as well as the action state and output of flexible adjustable resources.

[0165] Furthermore, in the real-time stage, the photovoltaic forecast data for the next hour is obtained. The distribution of photovoltaic forecast scenarios is updated and the weight factors are determined. The reference value of the day-ahead result is tracked and optimized in real time to obtain the photovoltaic absorption improvement margin and system tracking cost, as well as the action status and output of flexible and adjustable resources.

[0166] For example, the real-time stage dispatch is executed 1 hour in advance on the dispatch day with a time granularity of 15 minutes, using the day-ahead dispatch result as a reference. The photovoltaic short-term forecast data is input, and the output power of flexible resources such as ESS, Class I and Class II DR and power purchase is solved through rolling optimization.

[0167] The intraday dispatch model of the active distribution network is described below.

[0168] For example, the intraday scheduling model rolls over and tracks the day-ahead scheduling plan with the goal of minimizing the output deviation cost of flexible dispatchable resources, DR response compensation cost, and ESS depreciation cost. The objective function of the intraday scheduling model can be in the following form:

[0169]

[0170] Among them, the first three items in the square brackets of the above formula are the main grid power purchase cost, DR incentive cost and ESS life loss cost; ΔP t , p,t They are the deviation power and unit cost of the flexible resources participating in real-time scheduling. ΔP t It can be in the following forms:

[0171]

[0172] Among them, the variables with superscript “′” are the day-ahead scheduling results and are used as reference values ​​in the intraday scheduling stage; ΔP t It indicates the mismatch between the actual output and the planned value on the day before. When it is positive, it means that the actual output is greater than the plan on the day before, and there is a phenomenon of abandoned light in the distribution network. When it is negative, it means that the actual output is less than the plan on the day before, the distribution network is underpowered, and there is a risk of load shedding. The unit deviation power cost coefficient is used as a means for the distribution network to improve its absorption capacity. By setting different ρ p,t The penalty value is given to encourage distribution networks to give priority to the use of photovoltaic power.

[0173] The objective function of the intraday dispatch model is described above, and the constraints of the intraday dispatch model are described below. The constraints of the intraday dispatch model may include equipment ramp power constraints, demand response load adjustment constraints, and power balance constraints.

[0174] 1) Equipment climbing power constraint can be expressed as:

[0175]

[0176] Where: * represents the power of ES and other equipment involved in real-time scheduling.

[0177] 2) The demand response load adjustment constraint can be expressed as:

[0178]

[0179] In the formula, The loads participating in demand response represent four types of demand response loads.

[0180] 3) Power balance constraint. In the real-time stage, only active power is considered:

[0181]

[0182] The remaining constraints are consistent with the day-ahead constraints. The traditional distribution network adjustable robust optimization uses two-stage robust optimization to optimize some decision variables, such as unit start and stop state variables, before the uncertain variables are realized. These decision variables have become non-adjustable hard constraints when random variable optimization problems are subsequently introduced. The solution process optimizes adjustable decision variables such as unit power within the entire adjustable range of the hard constraint set.

[0183] The above embodiments illustrate the objective function and constraints of the intraday scheduling model. The following embodiments illustrate the rolling optimization solution of the intraday scheduling model based on the MPC algorithm.

[0184] The core of the optimization method based on the MPC algorithm lies in rolling optimization and feedback correction. Assuming that the total number of MPC scheduling periods is H, in the h period, the future time domain is divided into the prediction time domain and the control time domain. The state value of the h-1 period is used as known information, and the optimization model is constructed by predicting the random variable information in [h, h+N-1] for solution. Among the N-1 period results obtained, only the result of the h+1 period is transmitted to the active distribution network system for execution.

[0185] Repeat the above operation until the scheduling optimization within H is completed. Compared with the day-ahead optimization, the MPC real-time optimization rolling process continuously updates the prediction of random quantities in the future time domain, and the online optimization obtains a solution with better timeliness.

[0186] In a possible implementation, after the active distribution network is dispatched and controlled based on the day-ahead dispatch result, the following is further included:

[0187] Step 401: In a current period of the intraday phase, obtain the actual output power of the adjustable resources in the current period.

[0188] Exemplarily, in the current period, the actual output power of the adjustable resource can be obtained through actual measurement.

[0189] Step 402: Obtain the planned output power of the adjustable resources in the current period and the future period in the day-ahead scheduling result.

[0190] For example, the day-ahead dispatch result may include the planned output power of the adjustable resources in each period of the dispatch day. Accordingly, in the day-ahead stage, the adjustable resources of the active distribution network have been configured based on the day-ahead dispatch result. Due to the difference between the prediction and the actual situation, the configuration based only on the day-ahead stage cannot fully meet the real-time demand within the day, so it is necessary to make real-time adjustments during the day.

[0191] Step 403: Based on the historical photovoltaic output data of the current time period and before the current time period, the photovoltaic output power of the future time period is predicted by the SARIMA model with a preset confidence.

[0192] Exemplarily, the SARIMA model used here may be the same as the SARIMA model in step 302. Further, the data based on which the prediction is made in step 403 may be real-time updated data, for example, historical photovoltaic output data for the current period and before the current period. Further, the preset confidence used here may be the same as or different from the preset confidence in step 302. The preset confidence in step 403 may be set based on the uncertainty requirements of the uncertainty set. This will be described in the subsequent embodiments.

[0193] Step 404: Based on the actual output power and the planned output power, obtain the deviation power of the adjustable resources in the current period.

[0194] Step 405: Based on the predicted photovoltaic output power and the planned output power, the deviation power of the adjustable resources in the future period is obtained.

[0195] Step 406: Based on the deviation power of the adjustable resources in the current time period and the future time period, the intraday scheduling model of the active distribution network is optimized and solved with the goal of minimizing the scheduling cost of the adjustable resources and the deviation power scheduling cost of the active distribution network in multiple future time periods, and the day-ahead scheduling results of the active distribution network in multiple future time periods are obtained.

[0196] Step 407: In the first time period among multiple time periods in the future, the active distribution network is dispatched and controlled based on the day-ahead dispatch result of the first time period; and the first time period is used as the current time period of the intraday stage, and the day-ahead dispatch result is rolled optimized through the intraday dispatch model.

[0197] Exemplarily, the first time period is used as the current time period of the intraday phase, and steps 401 to 406 are repeatedly executed to perform rolling optimization on the day-ahead scheduling result.

[0198] Considering the possible errors in the real-time updated prediction information, the following embodiment improves the MPC optimization algorithm based on the affine idea. The improved affine MPC optimization model is as follows:

[0199]

[0200] Where: p j are the affine formula, expected value and error of the random quantity of the jth node respectively; is the affine decision vector for the h-period iteration; the constraint g l and h m They are equality constraints and inequality constraints respectively.

[0201] Based on the above affine MPC optimization model, the objective function can be transformed into an affine equivalent. After affine transformation, the objective function includes two parts: the target expected value and the affine radius corresponding to the SARIMA confidence level.

[0202] In one possible implementation, the objective function of the intraday scheduling model is:

[0203]

[0204] F t，j =ρ p,t |ΔP t |

[0205] Where w represents the weight coefficient, w∈[0,1]; F t,0 represents the expected value of the target cost in period t, or called the target expected value; F t,j The error term for the cost in period t; ΔP t represents the deviation power of adjustable resources; ρ p,t Represents the unit adjustment cost of adjustable resources.

[0206]

[0207] Compared with the objective function of the day-ahead dispatch model: the main grid payment cost, backup power generation cost, DR incentive cost and ES loss cost correspond to the target expected value, and the deviation power cost of the flexible resource corresponds to the affine radius cost caused by the error.

[0208] The affine radius of the optimization target is affected by the random power fluctuation of photovoltaics and represents the target solution set space of the uncertainty optimization problem. The larger the value, the closer the problem is to uncertainty optimization and the less conservative it is. Conversely, the smaller the value, the more conservative it is.

[0209] Therefore, if the affine MPC method is used to make the optimization scheme have stronger random power absorption capability, that is, less conservative optimization, the affine radius target F caused by the error must be minimized as much as possible. t,j The weight factor w is used to adjust the constraint relationship between the target expected value and the affine radius, and an online adjustable affine MPC optimization model can be obtained.

[0210] Target affine radius F t,j Involving the online calculation of multiple power, it needs to absorb the photovoltaic forecast information of the next N time periods, which increases the amount of calculation required for solution execution. In order to speed up the optimization algorithm, the target solution is relaxed as follows:

[0211]

[0212] ΔP t，+ ≥0,ΔP t，≥ 0

[0213] The optimization of problem solving is to minimize the objective function, ΔP t，+ , ΔP t，- The non-negativity of ensures the effectiveness of the target penalty factor. The confidence interval of the SARIMA forecast data at time h provides the error boundary of the random power, that is, the affine radius of the random variable. Therefore, the robust optimization of the affine MPC intraday scheduling model executed within the confidence interval narrows the error value range, avoids blind optimization of random paths in a wider fluctuation range, and accelerates the solution speed.

[0214] Exemplarily, the objective function of the affine MPC optimization model based on SARIMA prediction can also be summarized as:

[0215]

[0216] In a possible implementation, before optimizing and solving the intraday dispatch model of the active distribution network, it also includes: obtaining the confidence of the SARIMA model, wherein the confidence is positively correlated with the target conservatism of the intraday dispatch model of the active distribution network. Based on the confidence of the SARIMA model, the error range of the photovoltaic output is determined. Based on the error range of the photovoltaic output, the error range of the dispatch cost of the intraday dispatch model of the active distribution network is determined as the affine radius. Based on the affine radius of the dispatch cost and the expected value of the target cost, the objective function of the intraday dispatch model of the active distribution network is obtained.

[0217] In the embodiment of the present invention, the confidence interval of the SARIMA prediction data at a certain moment provides the error boundary of the random power, that is, the affine radius of the random variable. In the embodiment of the present invention, the conservatism of the robust solution result of the intraday scheduling model can be adjusted by adjusting the confidence of the SARIMA model, thereby realizing the adjustable conservatism of robust optimization. Adjusting the confidence by SARIMA can narrow the error value range, avoid blind optimization in random paths with a wide fluctuation range, and accelerate the solution speed.

[0218] Figure 7 is a real-time affine MPC scheduling optimization flow chart provided by an embodiment of the present invention. Figure 7, initialize \(h = 1\); obtain the component state variables and power variables at the \((h - 1)\)th moment, the charge information of the ESS, and perform affine optimization.

[0219] The affine optimization includes: obtaining the stochastic power prediction information for the \(h\) to \(h - N + 1\) time periods by SARMA prediction; inputting the confidence interval of the SARMA prediction to determine the MPC affine radius. Optimize the solution objective to obtain the scheduling results for the next \(N\) time periods.

[0220] After the affine optimization, only issue and execute the scheduling execution power for the \(h\)th time period. If \(h < T\), it means the program has not ended. Continue to set \(h = h + 1\), repeat obtaining the component state variables and power variables at the \((h - 1)\)th moment, the charge information of the ESS, and perform affine optimization.

[0221] The embodiment of the present invention constructs an affine MPC rolling optimization model based on the SARIMA prediction of the expected curve and confidence interval of photovoltaic power. The cost in the real - time stage is measured by a weight factor for the system operation cost interval corresponding to the photovoltaic prediction confidence interval. An increase in the value of the weight factor indicates that the operation plan is closer to deterministic optimization, which is in line with the improvement of the photovoltaic prediction confidence level.

[0222] The multi - time - scale affine optimization method proposed in the embodiment of the present invention can fully exploit the distribution characteristics of distributed photovoltaic historical data, improve the ability of the optimization scheduling scheme to cope with uncertain factors, reduce the economic cost of system operation, and enhance the response potential of DR loads.

[0223] It should be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0224] The following is the device embodiment of the present invention. For the details not described in detail, reference can be made to the corresponding method embodiment above.

[0225] Figure 8 The structural schematic diagram of the robust optimization scheduling device for the active distribution network provided by the embodiment of the present invention is shown. For the sake of convenience of description, only the parts related to the embodiment of the present invention are shown and are described in detail as follows:

[0226] As Figure 8As shown, a robust optimization scheduling device 8 for an active distribution network includes: an acquisition module 81, which is used to obtain historical photovoltaic output data of distributed photovoltaics in the active distribution network. A prediction module 82, which is used to predict photovoltaic output data based on historical photovoltaic output data through a SARIMA model with a preset confidence. An uncertainty set acquisition module 83, which is used to determine the photovoltaic output distribution range based on the predicted photovoltaic output data, and use it as an uncertainty set of photovoltaic output, wherein the preset confidence represents the degree of uncertainty of the uncertainty set. A solution module 84, which is used to perform robust optimization solution on the day-ahead scheduling model of the active distribution network based on the uncertainty set of photovoltaic output and with the goal of minimizing the scheduling cost of the adjustable resources of the active distribution network, to obtain the day-ahead scheduling result of the active distribution network; wherein the day-ahead scheduling result includes the planned output power of the adjustable resources. A scheduling module 85, which is used to perform scheduling control on the active distribution network based on the day-ahead scheduling result.

[0227] The embodiment of the present invention predicts photovoltaic output data through the SARIMA model, determines the distribution range of photovoltaic output as an uncertainty set, and represents the degree of uncertainty of the uncertainty set through the confidence of the SARIMA model. Furthermore, based on the uncertainty set of the improved photovoltaic output, robust optimization is solved, the day-ahead scheduling result is obtained, and scheduling control is performed. Due to the improvement, the probability of photovoltaic output appearing within the distribution range of the uncertainty set is greater, and the probability of appearing outside the distribution range of the uncertainty set is smaller, which avoids blindly searching for advantages and disadvantages in the box interval during robust optimization, thereby reducing the conservatism of the robust optimization scheme. At the same time, the range of the uncertainty set is more in line with the actual photovoltaic output scenario, the range of the uncertainty set is reduced, and the efficiency of robust optimization solution is improved.

[0228] In one possible implementation, the uncertainty set obtaining module is specifically used to: for any prediction period, determine the extreme power point of the photovoltaic output data within the prediction period; based on the extreme power point of the photovoltaic output data in each period, obtain the envelope of the photovoltaic output data, wherein the envelope includes a maximum envelope and a minimum envelope; determine the range between the maximum envelope and the minimum envelope as the photovoltaic output distribution range; and use the photovoltaic output distribution range as the uncertainty set of the photovoltaic output.

[0229] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0230] Those of ordinary skill in the art will appreciate that the templates, units, and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0231] If the module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned active distribution network robust optimization scheduling method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. Computer-readable media may include: any entity or device that can carry computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier signal, telecommunication signal and software distribution medium.

[0232] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A method for robust optimization scheduling of active distribution network, characterized in that: include: Obtain historical photovoltaic output data of distributed photovoltaics in active distribution networks; Based on the historical photovoltaic output data, the photovoltaic output data is predicted by a SARIMA model with a preset confidence level; Determine a photovoltaic output distribution range based on the predicted photovoltaic output data and use it as an uncertainty set of photovoltaic output, wherein the preset confidence represents the degree of uncertainty of the uncertainty set; Based on the uncertain set of photovoltaic output, with the goal of minimizing the dispatching cost of the adjustable resources of the active distribution network, the day-ahead dispatching model of the active distribution network is robustly optimized and solved to obtain the day-ahead dispatching result of the active distribution network; wherein the day-ahead dispatching result includes the planned output power of the adjustable resources; Based on the day-ahead dispatch result, the active distribution network is dispatched and controlled.

2. The method for robust optimization and dispatching of active distribution network according to claim 1, characterized in that: The photovoltaic output distribution range is determined based on the predicted photovoltaic output data, and the photovoltaic output uncertainty set includes: For any forecast period, determining the extreme power point of the photovoltaic output data within the forecast period; Based on the extreme power points of the photovoltaic output data in each time period, an envelope of the photovoltaic output data is obtained, wherein the envelope includes a maximum envelope and a minimum envelope; The range between the maximum envelope line and the minimum envelope line is determined as the photovoltaic output distribution range; The photovoltaic output distribution range is regarded as an uncertain set of photovoltaic output.

3. The active distribution network robust optimization scheduling method according to claim 2 is characterized in that: The uncertainty set of photovoltaic output is: Among them, Ω pv represents the uncertainty set of PV output; represents the expected value of the photovoltaic output at node j during period t; represents the fluctuation value of photovoltaic output at node j during period t; ρ j,t represents the confidence of the SARIMA model at node j in period t; θ is the uncertainty adjustment coefficient of distributed photovoltaics, which represents the upper limit of the sum of the confidences of all nodes of distributed photovoltaics.

4. The method for robust optimization and dispatching of active distribution network according to claim 1, characterized in that: Based on the uncertain set of photovoltaic output, the day-ahead dispatch model of the active distribution network is robustly optimized to minimize the dispatch cost of the adjustable resources of the active distribution network. The day-ahead dispatch results of the active distribution network include: Based on the uncertainty set of the photovoltaic output and the affine relationship between the photovoltaic output and the adjustable decision variables of the day-ahead dispatch model of the active distribution network, an affine uncertainty set of the adjustable decision variables is obtained; Based on the affine relationship between the photovoltaic output and the adjustable decision variable, and the day-ahead dispatch model of the active distribution network, a robust optimization dispatch model is obtained; Based on the uncertainty set of the photovoltaic output and the affine uncertainty set of the adjustable decision variables, the robust optimization scheduling model is dually solved with the goal of minimizing the scheduling cost of the adjustable resources of the active distribution network to obtain the day-ahead scheduling result of the active distribution network.

5. The method for robust optimization and dispatching of active distribution network according to claim 4, characterized in that: The affine relationship between the photovoltaic output and the adjustable decision variable is: in, represents an adjustable decision variable; represents the initial value of the adjustable decision variable; f j (t) represents the preset affine proportional relationship; Indicates photovoltaic output.

6. The method for robust optimization and dispatching of active distribution network according to claim 1, characterized in that: After the dispatching control of the active distribution network is performed based on the day-ahead dispatching result, the method further includes: In the current period of the intraday stage, the actual output power of the adjustable resources in the current period is obtained; Obtain the planned output power of adjustable resources in the current period and future period in the day-ahead scheduling results; Based on the historical photovoltaic output data of the current period and before the current period, the photovoltaic output power of the future period is predicted by the SARIMA model with a preset confidence level; Based on the actual output power and the planned output power, obtaining the deviation power of the adjustable resources in the current period; Based on the predicted photovoltaic output power and the planned output power, the deviation power of the adjustable resources in the future period is obtained; Based on the deviation power of adjustable resources in the current period and future periods, the scheduling cost of adjustable resources and the deviation power scheduling cost of the active distribution network in multiple future periods are minimized, and the day-ahead scheduling results of the active distribution network in multiple future periods are obtained. In the first time period of multiple time periods in the future, the active distribution network is dispatched and controlled based on the day-ahead dispatch result of the first time period; and the first time period is used as the current time period of the intraday stage, and the day-ahead dispatch result is rolled optimized through the intraday dispatch model.

7. The method for robust optimization and dispatching of active distribution network according to claim 6, characterized in that: Before optimizing and solving the intraday dispatching model of the active distribution network, the method further includes: Obtaining the confidence of the SARIMA model, wherein the confidence is positively correlated with the target conservatism of the intraday dispatching model of the active distribution network; Based on the confidence of the SARIMA model, the error range of photovoltaic output is determined; Based on the error range of photovoltaic output, the error range of the dispatch cost of the intraday dispatch model of the active distribution network is determined as the affine radius; Based on the affine radius of the dispatch cost and the expected value of the target cost, the objective function of the intraday dispatch model of the active distribution network is obtained.

8. The method for robust optimization and dispatching of active distribution network according to claim 6, characterized in that: The objective function of the intraday scheduling model is: F t,j =ρ p,t |ΔP t | Where w represents the weight coefficient, w∈[0,1]; F t,0 represents the expected value of the target cost in period t; F t,j The error term for the cost in period t; ΔP t represents the deviation power of adjustable resources; ρ p,t Represents the unit adjustment cost of adjustable resources.

9. An active distribution network robust optimization dispatching device, characterized in that: include: An acquisition module is used to obtain historical photovoltaic output data of distributed photovoltaics in the active distribution network; A prediction module, configured to predict the photovoltaic output data based on the historical photovoltaic output data by using a SARIMA model with a preset confidence level; An uncertainty set obtaining module, used for determining a photovoltaic output distribution range based on the predicted photovoltaic output data, and taking the photovoltaic output as an uncertainty set, wherein the preset confidence represents the uncertainty degree of the uncertainty set; A solution module is used for performing robust optimization solution on the day-ahead scheduling model of the active distribution network based on the uncertainty set of the photovoltaic output and taking the scheduling cost of the adjustable resources of the active distribution network as the minimum, so as to obtain the day-ahead scheduling result of the active distribution network; wherein the day-ahead scheduling result includes the planned output power of the adjustable resources; A dispatching module is used to dispatch and control the active distribution network based on the day-ahead dispatching result.

10. The active distribution network robust optimization dispatching device according to claim 9, characterized in that: The uncertain set obtaining module is specifically used for: For any forecast period, determining the extreme power point of the photovoltaic output data within the forecast period; Based on the extreme power points of the photovoltaic output data in each time period, an envelope of the photovoltaic output data is obtained, wherein the envelope includes a maximum envelope and a minimum envelope; The range between the maximum envelope line and the minimum envelope line is determined as the photovoltaic output distribution range; The photovoltaic output distribution range is regarded as an uncertain set of photovoltaic output.

Images