Robust optimal dispatching method for distribution network based on line loss and renewable energy output uncertainty

Abstract

The present invention belongs to the field of electric power grid technology, and relates to a distribution network robust optimization scheduling method based on line loss and uncertainty of new energy output, which includes: S1 obtaining distribution network historical data, and obtaining distribution network load curve and new energy output interval through a prediction model; S2 establishing a distribution network double-layer robust optimization scheduling model with the goal of minimizing distribution network line loss and new energy output uncertainty; S3 converting the double-layer robust optimization scheduling model into a mixed integer linear programming model through KKT optimality conditions and linear programming duality theorem; S4 solving the mixed integer linear programming model to obtain load incentive settings and scheduling schemes, and completing distribution network robust optimization scheduling control. The present invention is applicable to the problem of reducing line loss in distribution network operation and maintenance, especially in scenarios where new energy output accounts for a high proportion, adjusting load usage in real time, reducing load peak-to-valley differences, resisting uncertainty in new energy output, and improving the power supply reliability of the distribution network.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric power grids, and in particular relates to a robust optimization scheduling method for a distribution network based on line losses and uncertainty in the output of renewable energy. Background Art

[0002] In traditional power systems, excessive line losses are a common problem during power transmission, directly impacting the economic viability and operational efficiency of the power system. With the expansion of power grids and the increase in power demand, effectively accounting for and reducing line losses in dispatch has become a hot topic. Generally, line losses increase with increasing transmission distances and load fluctuations. Therefore, in optimizing power system dispatch, it is necessary to incorporate line losses into power transmission planning to ensure the practical implementation of dispatch plans.

[0003] With the large-scale integration of renewable energy, uncertainty in the power system has increased significantly. Renewable energy sources exhibit significant volatility and uncertainty, with their output fluctuating with environmental factors, making them difficult to accurately predict. Traditional dispatching methods struggle to adapt to these fluctuations, often resulting in output failing to meet load demand or causing grid imbalance. Therefore, to ensure stable system operation and mitigate uncertainty risks, it is necessary to incorporate the uncertainty of renewable energy sources into dispatching models.

[0004] Robust optimization is a common uncertainty optimization method widely used in optimization problems under uncertainty, particularly in power system scheduling with high rates of renewable energy integration. By introducing robust optimization methods, scheduling plans can be optimized under uncertain renewable energy output conditions, reducing scheduling deviations caused by volatility. By accounting for line losses and renewable energy output uncertainty, robust optimization methods can improve the stability and feasibility of scheduling plans, thereby ensuring efficient operation and power supply reliability of the power system.

[0005] As the challenges posed by line losses and renewable energy output uncertainty in power systems grow, traditional dispatch methods struggle to meet the safety, economy, and efficiency requirements of modern power systems. Therefore, developing a robust optimal dispatching method that incorporates line losses and renewable energy output uncertainty has significant practical significance and promising applications. This method will provide a reliable dispatching strategy for future power grids operating with a high proportion of renewable energy integration, improving the adaptability and risk mitigation capabilities of the power system. Summary of the Invention

[0006] In response to the shortcomings of the existing technology, the present invention provides a robust optimization scheduling method for distribution networks based on line losses and uncertainty in renewable energy output. The method is suitable for reducing line losses in distribution network operation and maintenance, especially in scenarios where the output of renewable energy accounts for a high proportion. It can adjust load usage in real time, reduce load peak-valley differences, resist the uncertainty of renewable energy output, and improve the power supply reliability of the distribution network.

[0007] To achieve the above objectives, the present invention discloses a robust optimization scheduling method for a distribution network based on line loss and uncertainty of renewable energy output, which includes:

[0008] S1: Obtain historical data of the distribution network and obtain the distribution network load curve and new energy output range through the prediction model;

[0009] S11: Obtain historical load incentive output L and new energy historical output data N of the low-voltage distribution network through the data acquisition and monitoring control system;

[0010] S12: Establish a historical load incentive output prediction model for low-voltage distribution network based on BP neural network;

[0011] S13: Establish a new energy historical output data forecasting model based on ARIMA time series;

[0012] S14: Input the historical load incentive output L of the low-voltage distribution network and the historical output data N of new energy into the prediction model to calculate the load incentive output P at the next time node ld and new energy output range [P ne- ,P ne+ ];

[0013] S2: Establish a two-layer robust optimization scheduling model for distribution network with the goal of minimizing distribution network line losses and the uncertainty of renewable energy output;

[0014] The minimization of the distribution network line loss is the first-level optimization model; the uncertainty of the new energy output is the second-level optimization model; the two-level robust optimization scheduling model of the distribution network is obtained as follows:

[0015]

[0016] Where: (i, j) is the line between the first node i and the second node j; i is the first node number; j is the second node number; r ij is the resistance of line (i, j); P ij,t is the active power of line (i, j) at time t; Q ij,t is the reactive power of line (i, j) at time t; (i, j)=>i is the set of lines flowing to the first node i; i=>(i, j) is the set of lines flowing out of the first node i; V i,t is the voltage of the first node i at time t; x ij is the reactance of line (i, j); V j,t is the voltage of the second node j at time t; P i ld+ is the upper limit of load incentive output of the first node i; P ild- is the lower limit of load excitation output of the first node i; P i ne+ is the upper limit of the new energy output of the first node i; P i ne- is the lower limit of the new energy output of the first node i; is the minimum optimization target of the load incentive output of the first node i; is the maximum optimization target of the new energy output of the first node i; t is the time parameter; T is the total number of time parameters; V i,t is the voltage of the first node i at time t; P i ne is the new energy output of the first node i; P i ld is the load excitation output of the first node i; is the load excitation reactive power of the first node i;

[0017] S3: The two-level robust optimization scheduling model in step S2 is transformed into a mixed integer linear programming model through the KKT optimality condition and the linear programming duality theorem:

[0018]

[0019] in: is the first dual variable of the first node i; is the first dual variable of the second node j; is the second dual variable of the first node i; is the second dual variable of the second node j; is the third dual variable of the node; is the first dual variable of the inequality at the first node i; is the second dual variable of the first node i inequality; is the dual variable Boolean variable corresponding to the constraint relaxation; is the dual variable Boolean variable corresponding to the constraint relaxation; θ i,t is the power factor angle of the first node i at time t; M is a constant parameter;

[0020] S4: Use the next time node load obtained in step S1 to stimulate the output P ld and new energy output range [P ne- ,P ne + ] is used as the constraint condition to solve the mixed integer linear programming model described in step S3, obtain the load incentive setting and scheduling plan for the next time node, and complete the robust optimization scheduling control of the distribution network.

[0021] Preferably, the historical load incentive output L prediction model of the low-voltage distribution network based on the BP neural network in step S12 includes an input layer, a hidden layer, and an error calculation and back propagation layer of the weight adjustment rule. The specific model is:

[0022]

[0023] Where: x i is the i-th input; w ij is the weight between the input layer and the hidden layer; b j is the input layer bias; f(·) is the activation function; h j is the output of the jth hidden node; v jk is the weight between the hidden layer and the output layer; c k is the output layer bias; g(·) is the output layer activation function; y k Output value of the kth BP neural network model; is the actual value of the kth BP neural network model; E is the loss function; m is the total number of second nodes; n is the total number of first nodes; j is the index of the hidden layer node; k is the index of the output layer node; p is the total number of output nodes.

[0024] Preferably, the prediction model of new energy historical output data N based on ARIMA time series in step S13 is specifically:

[0025]

[0026] Where: y t is the predicted value at time t; c is the constant parameter; is the coefficient of the first autoregressive term; is the coefficient of the second autoregressive term; is the coefficient of the pth autoregressive term; ε t is the white noise at time t; θ2 is the coefficient of the second moving average term; θ3 is the coefficient of the third moving average term; θ q is the coefficient of the qth moving average term.

[0027] Preferably, the first-level optimization model in step S2 takes minimizing the line loss of the distribution network as the optimization goal, specifically:

[0028]

[0029] Where: x i,t is the elastic coefficient of the first node i at time t; The load before implementing demand response; The load after implementing demand response; is the load incentive variable of node i at time t before implementing demand response; is the load incentive variable of node i at time t after implementing demand response; r peak is the load excitation peak value; r valley is the load excitation valley; T peak is the load excitation peak period; T valley It is the load incentive valley period; This is the upper limit after the implementation of demand-side response; It is the lower limit after implementing demand-side response.

[0030] Preferably, the second-level optimization model in step S2 determines the next node load curve data P according to the upper and lower limits of the new energy output peak and valley. ld Adjustment range [P ld- ,P ld+ ]for:

[0031]

[0032] in, is the load incentive upper limit of the first node i at time t; is the predicted load size of the first node i at time t; The lower limit of incentives after implementing demand-side response; To provide incentives before implementing demand-side response; x i,t Correction parameters for load excitation; is the lower limit of load excitation of the first node i at time t; It is the incentive cap after implementing demand-side response.

[0033] Preferably, the KKT optimality condition in step S3 is a constraint condition that allows the first-layer optimization model to be transformed into a second-layer optimization model through Lagrangian partial derivatives, specifically:

[0034]

[0035] By transforming the dual conditions, with the help of the dual variables The constraints are converted into:

[0036]

[0037] Preferably, the linear programming dual theorem in step S3 is specifically:

[0038] At the optimal solution, the objective function values ​​of the original problem and the dual problem are equal, which is the sum of the products of the dual variables and the constant terms in the original first-level optimization model. The equation is:

[0039]

[0040] Preferably, in step S4, the mixed integer linear programming model in step S3 is solved by a branch and bound algorithm, specifically:

[0041] S41: Node splitting, splitting the integer variables, dividing the search space, generating subproblems, and ensuring that each subproblem contains the solution space;

[0042] S42: Relaxation solution: For the generated subproblems, by relaxing the integer constraints, the problem is converted into a linear programming problem and solved, and the upper and lower bounds of the optimal solution of the subproblem are obtained;

[0043] S43: Optimality judgment, when checking whether the relaxed solution of the current node meets the optimality condition or the relaxed solution cannot improve the global optimal solution; if the condition is met, stop splitting and record the optimal solution;

[0044] S44: Pruning operation, directly pruning nodes that do not meet the constraints or whose objective function values ​​are worse than the current optimal solution, reducing the amount of calculation and improving the solution efficiency;

[0045] S45: Output the algorithm solution result, and output the optimal scheduling solution solved by the branch and bound algorithm. The key parameters include the node load scheduling power, the voltage level of each node, and the line power flow.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] (1) This invention can effectively reduce line losses during power transmission by establishing a robust optimization scheduling model with the goal of minimizing line losses. This optimization method takes into account various uncertainties in actual operation, ensuring that the optimal scheduling solution can be achieved even when the output of new energy sources fluctuates, thereby improving the economy and efficiency of the power system.

[0048] (2) The present invention uses a neural network or time series method to predict node load curves and new energy output intervals, and combines it with an electricity price incentive demand response model to adjust user loads in real time and enhance the distribution network's ability to withstand the uncertainty of new energy output. This method enables users to actively respond to changes in electricity prices, thereby balancing loads, reducing peak-to-valley load differences, and improving the power supply reliability of the distribution network, ensuring the stability of power supply when a high proportion of new energy is connected.

[0049] (3) The application of this invention in scenarios with a high proportion of new energy promotes the effective utilization of renewable energy. By optimizing the scheduling scheme, it can better integrate and utilize new energy resources such as wind power and photovoltaics, reduce dependence on traditional fossil energy, thereby promoting the development of green energy and achieving sustainable power supply. At the same time, the demand response mechanism based on electricity prices encourages users to use new energy at appropriate times, further improving the absorption capacity of renewable energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a flow chart of the robust optimization scheduling method for distribution networks based on line losses and uncertainty in renewable energy output according to the present invention;

[0051] Figure 2 The network structure of the distribution network in the embodiment of the present invention is a standard IEEE 33 node structure diagram;

[0052] Figure 3 This is the normalized graph of the load curve data of the distribution network and the new energy output interval of the present invention;

[0053] Figure 4 This is the worst new energy output curve obtained by the solution of the present invention;

[0054] Figure 5 The curves before and after the load scheduling of the distribution network of the present invention are shown in the figure;

[0055] Figure 6 This is a comparison diagram of line loss rates before and after load scheduling of the distribution network of the present invention. DETAILED DESCRIPTION

[0056] The exemplary embodiments, features, and aspects of the present invention will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise indicated.

[0057] The embodiment of the present invention provides a method for robust optimization scheduling of distribution network based on line loss and uncertainty of renewable energy output, such as Figure 1 As shown in the figure, historical data of the distribution network is obtained, and the distribution network load curve and renewable energy output range are obtained through the prediction model; a two-layer robust optimization scheduling model of the distribution network is established with the goal of minimizing distribution network line losses and renewable energy output uncertainty; the two-layer robust optimization scheduling model is transformed into a mixed integer linear programming model through the KKT optimality condition and the linear programming duality theorem; the mixed integer linear programming model is solved to obtain the load incentive setting and scheduling plan, and the robust optimization scheduling control of the distribution network is completed; it includes:

[0058] Step S1: Obtain historical data of the distribution network and obtain the distribution network load curve and new energy output range through the prediction model.

[0059] like Figure 2The distribution network structure used in this embodiment of the present invention is a standard IEEE 33-node low-voltage distribution system consisting of 33 grid nodes. Table 1 shows the locations and capacities of new energy access points, indicating the capacity of the corresponding nodes connected to the new energy source. This reflects the impact of new energy output uncertainty on the distribution network optimization process. New energy refers to renewable energy developed and utilized based on new technologies, including wind power, photovoltaic power generation, and tidal power generation.

[0060] Table 1 New energy access location and capacity

[0061] Node number New energy access capacity Node number New energy access capacity 3 250.54KVA 21 217.75KVA 6 264.47KVA 22 126.91KVA 7 932.15KVA 23 840.12KVA 13 708.69KVA 24 827.24KVA 17 521.22KVA 28 271.59KVA 18 246.16KVA 29 941.72KVA 19 109.70KVA 30 239.17KVA 20 217.75KVA 31 697.65KVA

[0062] Step S11: Obtain the historical load incentive output L and the historical new energy output data N of the low-voltage distribution network through the Supervisory Control and Data Acquisition (SCADA) system. The data time period can be set to one week or one month.

[0063] Step S12: Establish a historical load incentive output prediction model for the low-voltage distribution network based on a BP neural network, including an input layer, a hidden layer, and an error calculation and back propagation layer of weight adjustment rules. The specific model is:

[0064]

[0065] Where: x i is the i-th input; w ij is the weight between the input layer and the hidden layer; b j is the input layer bias; f(·) is the activation function; h j is the output of the jth hidden node; v jk is the weight between the hidden layer and the output layer; c k is the output layer bias; g(·) is the output layer activation function; y k Output value of the kth BP neural network model; is the actual value of the kth BP neural network model; E is the loss function; m is the total number of second nodes; n is the total number of first nodes; j is the index of the hidden layer node; k is the index of the output layer node; p is the total number of output nodes.

[0066] Step S13: Establish a new energy historical output data prediction model based on ARIMA time series, specifically:

[0067]

[0068] Where: y t is the predicted value at time t; c is the constant parameter; is the coefficient of the first autoregressive term; is the coefficient of the second autoregressive term; is the coefficient of the pth autoregressive term; ε t is the white noise at time t; θ2 is the coefficient of the second moving average term; θ3 is the coefficient of the third moving average term; θ q is the coefficient of the qth moving average term.

[0069] Step S14: Input the historical load incentive output L of the low-voltage distribution network and the historical output data N of new energy into the prediction model to calculate the load incentive output P at the next time node. ld and new energy output range [P ne- ,P ne+ ], the normalized results are as follows Figure 3 As shown in the figure, it shows the relationship between the load forecast curve and the output range of renewable energy during the operation of the distribution network, and depicts the upper and lower limits of the renewable energy output range.

[0070] Step S2: Establish a two-layer robust optimization scheduling model for the distribution network with the goal of minimizing distribution network line losses and uncertainty in renewable energy output.

[0071] Minimizing the line loss of the distribution network is the first-level optimization model, specifically:

[0072]

[0073] Where: x i,t is the elastic coefficient of the first node i at time t; The load before implementing demand response; The load after implementing demand response; is the load incentive variable of node i at time t before implementing demand response; is the load incentive variable of node i at time t after implementing demand response; r peak is the load excitation peak value; r valley is the load excitation valley; T peak is the load excitation peak period; T valley It is the load incentive valley period; This is the upper limit after the implementation of demand-side response; It is the lower limit after implementing demand-side response.

[0074] The uncertainty of new energy output is the second-level optimization model, which determines the next node load curve data P according to the upper and lower limits of new energy output peaks and valleys. ld Adjustment range [P ld- ,P ld+ ]for:

[0075]

[0076] in, is the load incentive upper limit of the first node i at time t; is the predicted load size of the first node i at time t; The lower limit of incentives after implementing demand-side response; To provide incentives before implementing demand-side response; x i,t Correction parameters for load excitation; is the lower limit of load excitation of the first node i at time t; It is the incentive cap after implementing demand-side response.

[0077] The two-layer robust optimization dispatch model of the distribution network is obtained as follows:

[0078]

[0079] Where: (i, j) is the line between the first node i and the second node j; i is the first node number; j is the second node number; r ij is the resistance of line (i, j); P ij,t is the active power of line (i, j) at time t; Q ij,t is the reactive power of line (i, j) at time t; (i, j)=>i is the set of lines flowing to the first node i; i=>(i, j) is the set of lines flowing out of the first node i; V i,t is the voltage of the first node i at time t; x ij is the reactance of line (i, j); V j,t is the voltage of the second node j at time t; P i ld+ is the upper limit of load incentive output of the first node i; P i ld- is the lower limit of load excitation output of the first node i; P i ne+ is the upper limit of the new energy output of the first node i; P i ne- is the lower limit of the new energy output of the first node i; is the minimum optimization target of the load incentive output of the first node i; is the maximum optimization target of the new energy output of the first node i; t is the time parameter; T is the total number of time parameters; V i,t is the voltage of the first node i at time t; P i ne is the new energy output of the first node i; P i ld is the load excitation output of the first node i; The reactive power is excited for the load of the first node i.

[0080] Step S3: Using the KKT optimality condition is to transform the first-level optimization model into the constraint condition of the second-level optimization model through Lagrangian partial derivative, specifically:

[0081]

[0082] By transforming the dual conditions, with the help of the dual variables The constraints are converted into:

[0083]

[0084] The use of the linear programming duality theorem makes the objective function values ​​of the original problem and the dual problem equal at the optimal solution, which is the sum of the products of the dual variables and the constant terms in the original first-level optimization model. The equation is:

[0085]

[0086] The two-level robust optimization scheduling model in step S2 is transformed into a mixed integer linear programming model through the KKT optimality condition and the linear programming duality theorem:

[0087]

[0088] in: is the first dual variable of the first node i; is the first dual variable of the second node j; is the second dual variable of the first node i; is the second dual variable of the second node j; is the third dual variable of the node; is the first dual variable of the inequality at the first node i; is the second dual variable of the first node i inequality; is the dual variable Boolean variable corresponding to the constraint relaxation; is the dual variable Boolean variable corresponding to the constraint relaxation; θ i,t is the power factor angle of the first node i at time t; M is a constant parameter.

[0089] Step S4: Use the next time node load obtained in step S1 to stimulate the output P ld and new energy output range [P ne- ,P ne+ ] is the constraint condition to solve the mixed integer linear programming model in step S3 by using the branch and bound algorithm, specifically:

[0090] Step S41: Node splitting, splitting the integer variables, dividing the search space, generating sub-problems, and ensuring that each sub-problem contains a solution space.

[0091] Step S42: Relaxation solution: for the generated sub-problems, the problem is converted into a linear programming problem by relaxing the integer constraints and solving it to obtain the upper and lower bounds of the optimal solution of the sub-problems.

[0092] Step S43: Optimality judgment, when checking whether the relaxed solution of the current node meets the optimality condition or the relaxed solution cannot improve the global optimal solution; if the condition is met, stop splitting and record the optimal solution.

[0093] Step S44: Pruning operation, directly pruning the nodes that do not meet the constraints or whose objective function values ​​are worse than the current optimal solution, reducing the amount of calculation and improving the solution efficiency.

[0094] Step S45: Output the algorithm solution, including the optimal dispatching solution obtained by the branch-and-bound algorithm. Key parameters include node load dispatching power, node voltage levels, and line power flow. The node load incentive settings and dispatching solution for the next time point are obtained, completing robust optimal dispatching control of the distribution network.

[0095] like Figure 4 The figure shows the worst-case scenario for renewable energy output, as calculated by the present invention. This scenario is generated by the present invention based on extreme assumptions about the uncertainty of renewable energy output. It demonstrates that regardless of how renewable energy output changes, the resulting load incentive settings and scheduling scheme for the next time node will result in a scheduling optimization result that is superior to that in the worst-case scenario. In other words, the model's calculated result is the most conservative solution, and actual conditions can even perform better than the calculated result.

[0096] like Figure 5 The curves before and after the load dispatch of the distribution network of the present invention are shown. They correspond to the curve of renewable energy output under the worst scenario, indicating that under this load incentive and dispatch result, the dispatch strategy under renewable energy output under the worst scenario is the best.

[0097] Finally, by scheduling the load before day, the impact of new energy access and its output uncertainty on line loss optimization is reduced. The line loss rate before and after load scheduling is compared. Figure 6 The total antenna loss rate before load scheduling is 3.9378%, and the total antenna loss rate after load scheduling is 3.7659%. The loss reduction effect is obvious, and the worst new energy output scenario is taken into account, so it has strong robustness.

[0098] The beneficial effects of the present invention are as follows: the present invention provides a robust optimization scheduling method for distribution networks based on line loss and uncertainty in renewable energy output. By establishing a robust optimization scheduling model with minimum line loss as the optimization goal, the line loss in the power transmission process can be effectively reduced. By using a neural network or time series method to predict the node load curve and the new renewable energy output range, combined with an incentive-based demand response model, the user load can be adjusted in real time, and the distribution network's ability to resist the uncertainty of renewable energy output can be enhanced. The embodiment of the present invention takes into account various uncertain factors in actual operation, ensuring that the optimal scheduling scheme can still be achieved under the condition of fluctuations in the output of new renewable energy, thereby improving the economy and efficiency of the power system; enabling users to actively respond to incentive changes, thereby balancing the load, reducing the peak-to-valley difference in load, improving the power supply reliability of the distribution network, and ensuring the stability of power supply when a high proportion of renewable energy is connected. By optimizing the scheduling scheme, new energy resources such as wind power and photovoltaics can be better integrated and utilized, reducing dependence on traditional fossil energy, thereby promoting the development of green energy and achieving sustainable power supply.

[0099] The embodiments described above are merely descriptions of preferred implementations of the present invention and are not intended to limit the scope of the present invention. Without departing from the spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should fall within the scope of protection determined by the claims of the present invention.

Claims

1. A robust optimization scheduling method for distribution networks based on line losses and uncertainty of renewable energy output, characterized in that: It includes: S1: Obtain historical data of the distribution network and obtain the distribution network load curve and new energy output range through the prediction model; S11: Obtain historical load incentive output L and new energy historical output data N of the low-voltage distribution network through the data acquisition and monitoring control system; S12: Establish a historical load incentive output prediction model for low-voltage distribution network based on BP neural network; S13: Establish a new energy historical output data forecasting model based on ARIMA time series; S14: Input the historical load incentive output L of the low-voltage distribution network and the historical output data N of new energy into the prediction model to calculate the load incentive output P at the next time node ld and new energy output range [P ne- ,P ne+ ]; S2: Establish a two-layer robust optimization scheduling model for distribution network with the goal of minimizing distribution network line losses and the uncertainty of renewable energy output; The minimization of the distribution network line loss is the first-level optimization model; the uncertainty of the new energy output is the second-level optimization model; the two-level robust optimization scheduling model of the distribution network is obtained as follows: Where: (i, j) is the line between the first node i and the second node j; i is the first node number; j is the second node number; r ij is the resistance of line (i, j); P ij,t is the active power of line (i, j) at time t; Q ij,t is the reactive power of line (i, j) at time t; (i, j)=>i is the set of lines flowing to the first node i; i=>(i, j) is the set of lines flowing out of the first node i; V i,t is the voltage of the first node i at time t; x ij is the reactance of line (i, j); V j,t is the voltage of the second node j at time t; P i ld+ is the upper limit of load incentive output of the first node i; P i ld- is the lower limit of load excitation output of the first node i; P i ne+ is the upper limit of the new energy output of the first node i; P i ne- is the lower limit of the new energy output of the first node i; is the minimum optimization target of the load incentive output of the first node i; is the maximum optimization target of the new energy output of the first node i; t is the time parameter; T is the total number of time parameters; V i,t is the voltage of the first node i at time t; P i ne is the new energy output of the first node i; P i ld is the load excitation output of the first node i; is the load excitation reactive power of the first node i; The first-level optimization model in step S2 takes minimizing the line loss of the distribution network as the optimization goal, specifically: Where: i,t is the elastic coefficient of the first node i at time t; The load before implementing demand response; The load after implementing demand response; is the load incentive variable of node i at time t before implementing demand response; is the load incentive variable of node i at time t after implementing demand response; ρ peak is the load excitation peak value; ρ valley is the load excitation valley; T peak is the load excitation peak period; T valley It is the load incentive valley period; This is the upper limit after the implementation of demand-side response; It is the lower limit after implementing demand-side response; The second-level optimization model in step S2 determines the next node load curve data P according to the upper and lower limits of the new energy output peak and valley. ld The adjustment range [P ld- ,P ld+ ]for: in, is the load incentive upper limit of the first node i at time t; is the predicted load size of the first node i at time t; The lower limit of incentives after implementing demand-side response; To provide incentives before implementing demand-side response; i,t Correction parameters for load excitation; is the lower limit of load excitation of the first node i at time t; The incentive cap after implementing demand-side response; S3: The two-level robust optimization scheduling model in step S2 is transformed into a mixed integer linear programming model through the KKT optimality condition and the linear programming duality theorem: in: is the first dual variable of the first node i; is the first dual variable of the second node j; is the second dual variable of the first node i; is the second dual variable of the second node j; is the third dual variable of the node; is the first dual variable of the inequality at the first node i; is the second dual variable of the first node i inequality; is the dual variable Boolean variable corresponding to the constraint relaxation; is the dual variable Boolean variable corresponding to the constraint relaxation; θ i,t is the power factor angle of the first node i at time t; M is a constant parameter; S4: Use the next time node load obtained in step S1 to stimulate the output P ld and new energy output range [P ne- ,P ne+ ] is used as the constraint condition to solve the mixed integer linear programming model described in step S3, obtain the load incentive setting and scheduling plan for the next time node, and complete the robust optimization scheduling control of the distribution network.

2. The method for robust optimization and dispatching of distribution networks based on line losses and uncertainty in renewable energy output according to claim 1 is characterized in that: The historical load incentive output L prediction model of the low-voltage distribution network based on the BP neural network in step S12 includes an input layer, a hidden layer, and an error calculation and back propagation layer of the weight adjustment rule. The specific model is: Where: x i is the i-th input; w ij is the weight between the input layer and the hidden layer; b j is the input layer bias; f(·) is the activation function; h j is the output of the jth hidden node; v jk is the weight between the hidden layer and the output layer; c k is the output layer bias; g(·) is the output layer activation function; y k Output value of the kth BP neural network model; is the actual value of the kth BP neural network model; E is the loss function; m is the total number of second nodes; n is the total number of first nodes; j is the index of the hidden layer node; k is the index of the output layer node; p is the total number of output nodes.

3. The method for robust optimization and dispatching of distribution networks based on line losses and uncertainty in renewable energy output according to claim 1 is characterized in that: The prediction model of new energy historical output data N based on ARIMA time series in step S13 is specifically: Where: y t is the predicted value at time t; c is the constant parameter; is the coefficient of the first autoregressive term; is the coefficient of the second autoregressive term; is the coefficient of the pth autoregressive term; ε t is the white noise at time t; θ2 is the coefficient of the second moving average term; θ3 is the coefficient of the third moving average term; θ q is the coefficient of the qth moving average term.

4. The method for robust optimization and dispatching of distribution networks based on line losses and uncertainty in renewable energy output according to claim 1 is characterized in that: The KKT optimality condition in step S3 is a constraint condition that allows the first-level optimization model to be transformed into the second-level optimization model through Lagrangian partial derivatives, specifically: By transforming the dual conditions, with the help of the dual variables The constraints are converted into:

5. The method for robust optimization and dispatching of distribution networks based on line losses and uncertainty in renewable energy output according to claim 1 is characterized in that: The linear programming dual theorem in step S3 is specifically: At the optimal solution, the objective function values ​​of the original problem and the dual problem are equal, which is the sum of the products of the dual variables and the constant terms in the original first-level optimization model. The equation is:

6. The method for robust optimization and dispatching of distribution networks based on line losses and uncertainty in renewable energy output according to claim 1, characterized in that: In step S4, the mixed integer linear programming model described in step S3 is solved by a branch and bound algorithm, specifically: S41: Node splitting, splitting the integer variables, dividing the search space, generating subproblems, and ensuring that each subproblem contains the solution space; S42: Relaxation solution: For the generated subproblems, by relaxing the integer constraints, the problem is converted into a linear programming problem and solved, and the upper and lower bounds of the optimal solution of the subproblem are obtained; S43: Optimality judgment, when checking whether the relaxed solution of the current node meets the optimality condition or the relaxed solution cannot improve the global optimal solution; if the condition is met, stop splitting and record the optimal solution; S44: Pruning operation, directly pruning nodes that do not meet the constraints or whose objective function values ​​are worse than the current optimal solution, reducing the amount of calculation and improving the solution efficiency; S45: Output the algorithm solution result, and output the optimal scheduling solution solved by the branch and bound algorithm. The key parameters include the node load scheduling power, the voltage level of each node, and the line power flow.

Images