A planning optimization method for connecting distributed power supply to power distribution network

Abstract

The application discloses a kind of distributed power supply access distribution network planning optimization method, it is related to power system planning and optimization technical field, this method includes: obtaining distribution network topological structure, load data and distributed power supply historical output data;Typical operation scenario set is constructed using scene generation and scene reduction technology;The double-layer optimization model of planning-operation collaborative optimization is constructed, the upper layer is planning layer, to determine access location and capacity with the minimum sum of investment construction equivalent annual cost and operating cost as target, the lower layer is operation layer, to optimize dispatching strategy with the minimum sum of operating cost and voltage deviation as target;Mixed solution strategy combining particle swarm algorithm and interior point method is used to carry out iterative solution;Optimal scheme is carried out multidimensional evaluation of economy, safety, environmental protection.This application realizes planning and operation collaborative optimization, effectively cope with source load uncertainty, applicable to distribution network distributed power supply access planning.

Description

Technical Field

[0001] This invention relates to the field of power system planning and optimization technology, specifically a planning and optimization method for distributed generation access to the distribution network. Background Technology

[0002] With the deepening of the "dual carbon" goals and the accelerated construction of new power systems, the scale of distributed generation in distribution networks continues to expand. Distributed generation sources such as photovoltaic and wind power have significant advantages, including being clean and low-carbon, flexible in deployment, and having short construction cycles, making them an important pathway to promote energy transformation. However, with the rapid growth of distributed generation capacity, the planning issues for their integration into distribution networks are becoming increasingly prominent, posing new challenges to the safe, economical, and reliable operation of distribution networks.

[0003] Currently, the planning methods for integrating distributed generation into the distribution network mainly fall into the following technical categories:

[0004] (I) Mathematical Programming-Based Methods. These methods establish distribution network planning models containing distributed generation sources, with the objective functions of minimizing system active power loss, voltage deviation, or annual comprehensive cost. Common mathematical programming methods include linear programming, nonlinear programming, and mixed-integer programming. These methods use the location and capacity of distributed generation sources as decision variables, solving for the optimal planning scheme under conditions such as power flow constraints, voltage constraints, and capacity constraints.

[0005] (ii) Heuristic Algorithm-Based Methods. These methods employ intelligent optimization algorithms such as genetic algorithms, particle swarm optimization, and simulated annealing to solve the location and capacity determination problem for distributed power sources. For example, a particle swarm optimization algorithm can be constructed to calculate the global optimal solution, which is then used as the optimal access scheme for the distributed power source.

[0006] (III) Sensitivity Analysis-Based Methods. This type of method establishes a sensitivity coefficient matrix to analyze the impact of the location and capacity of distributed generation on operating parameters such as node voltage and line power flow in the distribution network, and determines the optimal access scheme accordingly.

[0007] (iv) Two-level planning method. This type of method constructs upper-level and lower-level optimization models to handle planning and decision-making problems at different levels. The upper-level model determines the access nodes and capacity of distributed power sources, while the lower-level model performs detailed optimization based on the results of the upper-level model.

[0008] However, all of the above-mentioned existing technical solutions have varying degrees of technical defects:

[0009] First, there is a disconnect between planning and operation. Most existing methods focus on static site selection and capacity planning, failing to fully consider the impact of distributed generation output fluctuations and load time-varying characteristics on distribution network operation. In actual operation, planned solutions may lead to problems such as voltage exceeding limits and line overloads, resulting in a significant deviation between planning results and actual operational needs.

[0010] Second, there is insufficient handling of uncertainties. The output of distributed power sources (especially photovoltaic and wind power) is significantly random and intermittent, and the load demand also exhibits strong uncertainty. However, existing methods handle uncertainty in a relatively simple way, lacking effective robust optimization or stochastic optimization mechanisms, and thus failing to accurately reflect actual operating scenarios.

[0011] Third, the balance between multiple objectives is insufficient. Distributed power generation involves multiple objectives, including economic efficiency (investment cost, grid loss cost, electricity purchase cost), security (voltage stability, line margin), and environmental protection (new energy absorption rate). Existing methods are inadequate in the comprehensive optimization of multiple objectives, often neglecting one aspect for another.

[0012] Fourth, they lack adaptability and scalability. Most existing methods are designed for specific scenarios. When the distribution network topology changes or the types of distributed power sources increase, they need to be remodeled and solved, lacking universality and scalability.

[0013] In view of the shortcomings of the existing technologies, there is an urgent need for a distributed power source access network planning and optimization method that can achieve coordinated optimization of planning and operation, effectively cope with uncertainties, and comprehensively balance multiple objectives. Summary of the Invention

[0014] The technical problem to be solved by this invention is to overcome the existing defects and provide a planning and optimization method for distributed power generation access to the distribution network. This method can achieve coordinated optimization of the location and capacity of distributed power generation access with the operation strategy of the distribution network, effectively cope with the uncertainty of distributed power generation output and load demand, comprehensively optimize multiple objectives such as economy, safety and environmental protection, improve the adaptability and robustness of the planning scheme, and effectively solve the problems in the background technology.

[0015] To achieve the above objectives, the present invention provides the following technical solution: a planning and optimization method for distributed power generation access to a distribution network, comprising the following steps:

[0016] S1: Obtain the topology parameters of the distribution network, load data of each node, candidate access point information of distributed power sources and their historical output data, and perform data preprocessing.

[0017] S2: Based on the historical output data of distributed power sources and the load data of each node, a typical operating scenario set is constructed using scenario generation and scenario reduction techniques. The typical operating scenario set is used to characterize the output uncertainty and load uncertainty of distributed power sources.

[0018] S3: Construct a two-layer optimization model for planning-operation coordinated optimization. The upper layer of the two-layer optimization model is the planning layer, which aims to minimize the sum of the annual value costs of distributed power generation investment and construction and the distribution network operation costs. The connection location and connection capacity of distributed power generation are the upper-layer decision variables. The distribution network operation cost is the expected value of the weighted sum of the operation costs under each typical operation scenario according to the probability weight of each scenario. The lower layer of the two-layer optimization model is the operation layer, which aims to minimize the sum of the operation costs under each typical operation scenario and the node voltage deviation penalty term. The distribution network operation scheduling strategy is the lower-layer decision variable.

[0019] S4: The two-layer optimization model is iteratively solved using a hybrid solution strategy. The upper planning layer uses the particle swarm algorithm to perform a global search and passes the searched planning scheme to the lower running layer. The lower running layer uses the interior point method to solve the optimal running strategy under the current planning scheme and feeds back the optimal running cost to the upper planning layer. The optimal planning scheme is obtained through iterative coupling between the upper and lower layers.

[0020] S5: Perform a multi-dimensional comprehensive evaluation of the optimal planning scheme, including economic evaluation, safety evaluation and environmental evaluation, and output the final planning scheme for distributed power source access to the distribution network that meets the evaluation requirements.

[0021] Furthermore, in step S1: the number of nodes in the distribution network, the line impedance parameters between nodes, the line admittance parameters, and the transformer parameters are obtained to form a basic topology database of the distribution network; the annual load curves of each node are obtained according to a preset time period, and are classified according to seasonal type to generate typical daily load data corresponding to each season; the set of candidate access points for distributed power sources is determined, and the time series output data of each candidate access point in the historical period is obtained; and the obtained data are normalized.

[0022] Furthermore, in step S2: based on the historical output data of the distributed power source and the load data of each node, a number of initial uncertainty scenarios containing output fluctuations and load fluctuations of the distributed power source are generated using the Monte Carlo simulation method; the initial uncertainty scenarios are reduced using the fast prior elimination method based on probability distance to obtain a set of typical operating scenarios after the number is reduced; and the occurrence probability of each typical operating scenario is calculated and assigned a corresponding probability weight.

[0023] Furthermore, in step S3: the objective function of the upper planning layer includes the investment and construction cost of distributed power sources in the form of equal annual values, and the network loss cost under each typical operating scenario and the cost of purchasing electricity from the upper-level grid are weighted and summed according to the probability weight of each scenario as the expected value of the distribution network operating cost; the upper planning layer also sets the upper limit constraint of the total access capacity of distributed power sources, the maximum access capacity constraint of each candidate access point, and the upper limit constraint of the penetration rate of distributed power sources.

[0024] Furthermore, in step S3: the upper-level decision variables include binary state variables characterizing whether each candidate access point is connected to a distributed power source, and continuous variables characterizing the distributed power source access capacity at each candidate access point; the upper limit constraint on the total access capacity of distributed power sources means that the sum of the access capacities of all candidate access points shall not exceed the upper limit of the total capacity of distributed power sources allowed to be connected to the distribution network.

[0025] Furthermore, in step S3: the lower-level operation layer constructs a corresponding operation optimization sub-model for each typical operation scenario. The objective function of the operation optimization sub-model is the sum of the operation cost and the node voltage deviation penalty term under each typical operation scenario. The operation cost includes the operation and maintenance cost of distributed power sources and the cost of purchasing electricity from the upper-level grid. The node voltage deviation penalty term is the product of the voltage deviation index of each node in the entire distribution network and the preset penalty weight coefficient. The constraints of the operation optimization sub-model include node power balance constraints, upper and lower limits of voltage amplitude of each node, upper limit of transmission power of each branch, and upper, lower, and upper limits of active power output of distributed power sources.

[0026] Furthermore, in step S3: the lower-level decision variables include the active power output scheduling value of each distributed power source, the reactive power output scheduling value of each distributed power source, the number of switching groups of the reactive power compensation device, and the tap position of the on-load tap-changing transformer; the voltage deviation index of each node in the entire distribution network refers to the sum of the squares of the relative deviations between the actual voltage amplitude of each node and the corresponding node voltage reference value.

[0027] Furthermore, in step S4: each particle is encoded as a planning scheme containing the access status and access capacity of each candidate access point; the planning scheme obtained in the current iteration is passed to the lower operation layer, and the interior point method is used to solve the operation optimization sub-model for each typical operation scenario to obtain the optimal operation cost and minimum voltage deviation under the current planning scheme; the optimal operation cost of each typical operation scenario obtained by the lower operation layer is weighted by probability and superimposed on the objective function of the upper planning layer to update the fitness value of the current particle, and update the individual optimal position and global optimal position of the particle swarm; it is determined whether the maximum number of iterations has been reached or whether the global optimal fitness value meets the convergence accuracy. If it meets the requirements, the global optimal planning scheme is output; if it does not meet the requirements, the iteration continues.

[0028] Furthermore, in step S4: for the non-convex AC power flow constraints in the running optimization sub-model, the running optimization sub-model is first transformed into a convex optimization model using the second-order cone programming relaxation method, and then the original dual interior point method is used to solve the transformed convex optimization model.

[0029] Furthermore, in step S5: calculate the comprehensive economic indicators of the optimal planning scheme within a set time limit, wherein the comprehensive economic indicators are the sum of distributed power generation investment costs, operation and maintenance costs, distribution network loss costs, and costs of purchasing electricity from the upper-level grid; calculate the safety indicators of the optimal planning scheme under various typical operating scenarios, wherein the safety indicators include the voltage qualification rate of each node and the maximum load rate of each branch line; calculate the renewable energy absorption rate and the corresponding carbon emission reduction of the optimal planning scheme as environmental protection indicators; determine whether the economic indicators, safety indicators, and environmental protection indicators all meet the preset threshold requirements. If all are met, output the current planning scheme as the final planning scheme. If any indicator does not meet the requirements, return to step S4 to iterate and solve again until a final planning scheme that meets all evaluation indicators is obtained.

[0030] Compared with the prior art, the beneficial effects of the present invention are as follows: This planning and optimization method for distributed power generation to the distribution network has the following advantages:

[0031] 1. Enhanced feasibility through coordinated planning and operation optimization. This invention constructs a two-layer optimization model integrating planning and operation by considering operational strategy optimization during the planning stage. This avoids the problem of disconnect between planning schemes and actual operation in traditional methods, resulting in planning schemes with higher practical operability and engineering application value.

[0032] 2. Effectively address uncertainties and enhance the robustness of the solution. This invention constructs a multi-scenario uncertainty model through Monte Carlo simulation and scenario reduction technology. It effectively represents the uncertainty of distributed power output and load demand with a limited number of typical operating scenarios, enabling the planning scheme to maintain good performance under different operating scenarios.

[0033] 3. Comprehensive consideration of multiple objectives to achieve system optimization. This invention comprehensively optimizes multiple objectives such as economy (investment cost, operating cost), safety (voltage quality, line margin) and environmental protection (new energy absorption rate), avoiding the problem of neglecting one aspect for another caused by optimizing a single objective.

[0034] 4. High solution efficiency and strong applicability. This invention adopts a hybrid solution strategy with upper and lower layers separated—the upper layer uses the particle swarm optimization algorithm for global search, and the lower layer uses the interior point method for accurate solution—which improves solution efficiency while ensuring solution quality and is applicable to distribution networks of different scales and topologies.

[0035] 5. Excellent ability to handle non-convex constraints. This invention addresses the non-convex AC power flow constraints in the running optimization sub-model by employing a second-order cone programming relaxation method to transform them into a convex optimization model, effectively reducing the difficulty of solving the problem and improving computational efficiency and convergence. Attached Figure Description

[0036] Figure 1 This is the overall flowchart of the present invention. Detailed Implementation

[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] Please see Figure 1 The present invention provides the following embodiments:

[0039] Example 1

[0040] This embodiment takes a real 10kV medium-voltage distribution network as an example to provide a detailed description of the planning and optimization method for distributed power source access to the distribution network provided by the present invention.

[0041] Step S1: Data Acquisition and Preprocessing

[0042] Obtain the topology parameters of this distribution network. This network comprises 33 nodes and 32 branches, with a system base voltage of 10kV and a total load of approximately 3.8MW. Obtain the line impedance parameters (including resistance R and reactance X), line admittance parameters, and transformer parameters (including turns ratio, short-circuit impedance, etc.) for each node, forming a basic distribution network topology database. The basic topology data of the distribution network can be stored and managed using the IEEE standard test system data format.

[0043] The annual load curves for each node are obtained by hour, containing load data for 8760 time points. The load data is then categorized by season (March to May for spring, June to August for summer, September to November for autumn, and December to February of the following year for winter). The average daily load curves within each season are then averaged to generate typical daily load data for each season, with each typical day containing load values ​​(per unit) for 24 time points.

[0044] A set of candidate access points for the distributed power supply is determined. In this embodiment, eight nodes—nodes 4, 8, 12, 17, 21, 25, 30, and 33—are selected as candidate access points for the distributed power supply. The selection principle for candidate access points is to prioritize nodes located in the middle to later sections of the feeder, with higher load density and weaker voltage support capabilities.

[0045] The time-series output data of each candidate access point over the past year is obtained. In this embodiment, the distributed power source type is photovoltaic power generation. The hourly output data (in kW) of each candidate access point over the past year is obtained. This data can be calculated by combining the historical operation records of the photovoltaic power station or typical meteorological year data with the photovoltaic output model.

[0046] The acquired data are normalized to map data of different dimensions to the [0,1] interval, thus eliminating the impact of dimensional differences on subsequent optimization calculations. The min-max normalization method can be used for normalization.

[0047] Step S2: Construct a set of typical operating scenarios

[0048] Based on historical power output data and load data of each node, Monte Carlo simulation is used to generate 2000 initial uncertainty scenarios that include fluctuations in distributed power output and load. The specific implementation of Monte Carlo simulation is as follows: based on historical power output and load data of distributed power sources, their respective probability distribution functions are fitted (in this embodiment, photovoltaic power output is fitted with a Beta distribution, and load with a normal distribution). Random sampling is then performed from each probability distribution function to form a complete set of initial scenarios.

[0049] A fast priori elimination method based on probabilistic distance is used to reduce the number of initial uncertain scenarios from the aforementioned 2000 scenarios. The basic principle of the fast priori elimination method is: calculate the probabilistic distance between each scenario, sequentially delete the scenario with the smallest probability weight among the scenarios with the smallest probability distance, and transfer its probability weight to the scenario with the closest distance, until the number of remaining scenarios reaches a preset target value. In this embodiment, the target number of scenarios to be reduced is set to 10.

[0050] After scenario reduction, 10 typical operating scenarios are obtained. The probability of occurrence of each typical operating scenario is calculated and assigned a corresponding probability weight. The output is a set of typical operating scenarios containing the 10 typical operating scenarios and their probability weights. The sum of the probability weights of each scenario is 1.

[0051] Step S3: Construct a two-layer optimization model

[0052] (a) Upper-level planning layer:

[0053] The objective function is to minimize the sum of the annualized investment and construction costs of distributed generation and the operating costs of the distribution network. The investment and construction costs of distributed generation are included in the annualized value.

[0054] The calculation method for equivalent annual cost is as follows: the initial investment cost of the distributed power source is discounted into an annually equivalent cost based on the economic life of the equipment and the benchmark discount rate. In this embodiment, the unit investment cost of the photovoltaic power generation equipment is set at 4.0 yuan / W, the economic life of the equipment is taken as 20 years, and the benchmark discount rate is taken as 10%. Equivalent annual cost = Initial investment cost × Capital recovery factor, Capital recovery factor = [i×(1+i)^n] / [(1+i)^n-1], where i is the benchmark discount rate and n is the economic life of the equipment. In this embodiment, the capital recovery factor is approximately 0.1175. The selection of this set of parameters is based on the fact that the typical technical and economic parameters of photovoltaic power generation equipment are common knowledge in the field, and the above values ​​reflect the current industry average level. Those skilled in the art can adjust them according to the actual equipment type and market conditions.

[0055] The distribution network operating cost is the expected value obtained by weighting the sum of the network loss cost and the cost of purchasing electricity from the upstream grid under each typical operating scenario, according to the probability weights of each scenario. The network loss cost is determined based on the power flow calculation results under each scenario, and the electricity purchase cost is determined based on the switching power of the distribution network root node and the real-time electricity price. In this embodiment, the real-time electricity price is determined according to the time-of-use pricing mechanism: the electricity price during peak hours (8:00-11:00, 17:00-20:00) is 0.8 yuan / kWh, the electricity price during normal hours (11:00-17:00, 20:00-22:00) is 0.5 yuan / kWh, and the electricity price during off-peak hours (22:00-8:00 the next day) is 0.2 yuan / kWh.

[0056] The upper-level decision variables include: 8 binary state variables characterizing whether each candidate access point is connected to the distributed power source, with a value of 1 indicating connection and a value of 0 indicating no connection; and 8 continuous variables characterizing the capacity of the distributed power source at each candidate access point, with values ​​ranging from 0 to the maximum capacity of a single point.

[0057] The upper planning layer sets the following constraints:

[0058] (1) Maximum limit of total access capacity of distributed power sources: The sum of the access capacity of all candidate access points shall not exceed the maximum limit of total access capacity of distributed power sources allowed by the distribution network. In this embodiment, it is taken as 2.5MW. The method for determining this upper limit is: based on the evaluation results of the absorption capacity of the distribution network, specifically determined by comprehensively considering the short-circuit capacity, voltage regulation capability and protection coordination requirements of the distribution network. Those skilled in the art can obtain it through simulation calculation.

[0059] (2) Maximum single-point access capacity constraint for each candidate access point: The access capacity of each candidate access point shall not exceed 0.8MW. The method for determining this upper limit is as follows: based on the short-circuit capacity and thermal stability limit of the access point, it is determined according to the engineering experience rule of "not exceeding 10% of the short-circuit capacity of the access point".

[0060] (3) Upper limit constraint on the penetration rate of distributed generation: The total capacity of distributed generation shall not exceed 60% of the total load of the distribution network. The penetration rate constraint is used to ensure that the distribution network can still maintain sufficient regulation capacity and safety margin after the distributed generation is connected.

[0061] (ii) Lower operating layer:

[0062] For each typical operating scenario, a corresponding operation optimization sub-model is constructed. The objective function of the operation optimization sub-model is to minimize the sum of the operating cost and the node voltage deviation after penalty weighting under each typical operating scenario.

[0063] Operating costs include the operation and maintenance costs of distributed generation and the cost of purchasing electricity from the upstream grid. Operation and maintenance costs are calculated as the product of the distributed generation's output and the unit operation and maintenance cost; in this embodiment, the unit operation and maintenance cost is taken as 0.02 yuan / kWh. The cost of purchasing electricity from the upstream grid is calculated as the product of the switching power of the distribution network root node and the real-time electricity price, which is determined according to the time-of-use pricing mechanism.

[0064] The node voltage deviation penalty is the product of the voltage deviation index of each node in the entire distribution network and the preset penalty weighting coefficient. The voltage deviation index of each node in the entire distribution network refers to the sum of the squares of the relative deviations between the actual voltage amplitude of each node and the corresponding node voltage reference value.

[0065] In this embodiment, the penalty weight coefficient is set to 100. The selection principle for this coefficient is to ensure that the magnitude of the voltage deviation term in the objective function is comparable to that of the operating cost term, thereby effectively constraining voltage quality during the optimization process. Specifically, the coefficient is multiplied by the typical magnitude of the voltage deviation at each node (usually between 0.001 and 0.01), resulting in a value on the same order of magnitude as the operating cost term (usually between 1 and 10). Fine-tuning is then performed through several preliminary tests. Those skilled in the art can make appropriate adjustments based on the scale of the distribution network (number of nodes, line length, etc.).

[0066] The lower-level decision variables include: the active power output scheduling value of each distributed power source, the reactive power output scheduling value of each distributed power source, the number of switching groups of reactive power compensation devices, and the tap position of the on-load tap-changing transformer.

[0067] The constraints for running the optimization sub-model include:

[0068] (1) Node power balance constraint: The sum of the injected active power and reactive power of each node is zero (considering load and network losses). This constraint is an equality constraint.

[0069] (2) Upper and lower limits of voltage amplitude at each node: The voltage amplitude at each node should be maintained between 0.95 times the rated voltage and 1.05 times the rated voltage. This constraint is an inequality constraint.

[0070] (3) Upper limit constraint on transmission power of each branch: The transmission power of each branch shall not exceed its thermal stability limit capacity. This constraint is an inequality constraint.

[0071] (4) Upper limit constraint of active power output, lower limit constraint of active power output, upper limit constraint of reactive power output, and lower limit constraint of reactive power output: The output of each distributed power source should be within its technically adjustable range. In this embodiment, the upper limit of active power output of the photovoltaic inverter is the installed capacity, and the lower limit is 0; the upper and lower limits of reactive power output are determined according to 0.8 times the rated capacity of the inverter, and the power factor is adjustable in the range of 0.95 (leading) to 0.95 (lagging).

[0072] Step S4: Hybrid Solution

[0073] The particle swarm optimization (PSO) algorithm is used to solve the upper-level planning layer. The basic principle of PSO is to initialize a swarm of random particles (each particle representing a candidate solution), update the particle velocity and position by tracking the optimal position of individual particles and the global optimal position, and iteratively search for the optimal solution. PSO is an existing technology, and its specific implementation is common knowledge in the field.

[0074] In this embodiment, the population size is set to 100, the maximum number of iterations is set to 300, and the inertia weight adopts a linear decreasing strategy from 0.9 to 0.4. The specific implementation of the linear decreasing inertia weight strategy is as follows: in each iteration, inertia weight = 0.9 - (0.9 - 0.4) × current iteration number / maximum iteration number. The selection of these parameters is based on the population size and maximum number of iterations being determined comprehensively according to the decision variable dimension of the problem (16 dimensions in this embodiment) and the size of the search space. The linear decreasing inertia weight strategy is a classic parameter configuration scheme in particle swarm optimization, effectively balancing global search capability and local exploitation capability.

[0075] Each particle is encoded as a planning scheme containing the access state and access capacity of each candidate access point, with an encoding length of 16 dimensions (8 binary state variables + 8 continuous capacity variables).

[0076] The planning scheme obtained in the current iteration is passed to the lower-level operation layer, and the interior-point method is used to solve the operation optimization sub-model for each typical operation scenario to obtain the optimal operation cost and minimum voltage deviation under the current planning scheme. The interior-point method is a numerical optimization algorithm in the existing technology. Its basic principle is to transform inequality constraints into equality constraints by introducing a barrier function, and then use Newton's method to iteratively solve the KKT conditions.

[0077] The optimal operating costs of each typical operating scenario obtained from the lower operating layer are weighted by probability and then superimposed onto the objective function of the upper planning layer. This updates the fitness value of the current particle and the individual optimal position and global optimal position of the particle swarm.

[0078] Determine whether the maximum number of iterations (300) has been reached or whether the globally optimal fitness value meets the convergence accuracy (10). -4 If the conditions are met, the globally optimal planning solution is output; otherwise, the iteration continues. The convergence accuracy is determined by the change in the globally optimal fitness value being less than 10 over 20 consecutive iterations. -4 The threshold is used to determine if a solution has converged, and this threshold ensures a reasonable balance between solution accuracy and computational efficiency.

[0079] For the non-convex AC power flow constraints in the running optimization sub-model, the second-order cone programming relaxation method is first used to transform the running optimization sub-model into a convex optimization model. Then, the primal-dual interior-point method is used to solve the transformed convex optimization model. The basic principle of the second-order cone programming relaxation method is to transform the nonlinear constraints in the AC power flow equations into second-order cone constraints through variable substitution and relaxation. Specifically, intermediate variables are introduced to represent the square terms of node voltages and branch currents, relaxing the original non-convex power flow equations into second-order cone constraints, thus transforming the original non-convex problem into a convex optimization problem, ensuring the global optimality and convergence of the solution. The second-order cone programming relaxation method is common knowledge in this field, and its specific implementation can be found in relevant literature.

[0080] Step S5: Multi-dimensional comprehensive evaluation and solution output

[0081] The comprehensive economic index of the optimal planning scheme within a set period is calculated. This index is the sum of the investment cost, operation and maintenance cost, distribution network loss cost, and cost of purchasing electricity from the upstream grid. In this embodiment, the set period is 20 years, which corresponds to the typical economic service life of photovoltaic power generation equipment and is a conventional value in this field. Among them, the investment cost is calculated based on the equipment purchase and installation costs, and the operation and maintenance cost is calculated at a fixed annual percentage (2% of the initial investment).

[0082] Calculate the safety indicators of the optimal planning scheme under various typical operating scenarios, including the voltage qualification rate of each node and the maximum load rate of each branch line. The voltage qualification rate refers to the percentage of nodes with voltage amplitudes within the range of [0.95pu, 1.05pu] out of the total number of nodes. The maximum load rate of the line refers to the maximum value of the ratio of the actual transmission power of each branch line to its thermal stability limit capacity.

[0083] The optimal planning scheme's renewable energy integration rate and corresponding carbon emission reduction are calculated as environmental indicators. The renewable energy integration rate refers to the ratio of actual power generation to available power from distributed generation. Carbon emission reduction refers to the reduction in carbon emissions after integrating distributed generation compared to a pure electricity purchase model.

[0084] The system determines whether the economic, safety, and environmental indicators all meet preset threshold requirements. In this embodiment, the economic indicator requires a total cost reduction of at least 5% compared to the baseline scheme (without distributed power access); the safety indicator requires a voltage qualification rate of at least 99% and a maximum line load rate of no more than 80%; and the environmental indicator requires a renewable energy absorption rate of at least 95%. The selection of these thresholds is based on the following criteria: a 5% improvement in economic efficiency represents the minimum acceptable return on investment for the project; a 99% voltage qualification rate is a basic quality requirement for distribution network operation; an 80% maximum line load rate represents a typical safety margin under the N-1 criterion; and a 95% renewable energy absorption rate is a basic requirement of the national energy authorities for the utilization rate of renewable energy generation.

[0085] If all requirements are met, the current planning scheme is output as the final planning scheme; if any indicator is not met, the process returns to step S4 to iterate and solve again (adjusting the parameters of the particle swarm algorithm, such as increasing the population size or the maximum number of iterations) until a final planning scheme that meets all evaluation indicators is obtained.

[0086] Implementation Results

[0087] After optimizing the above distribution network using the method of this embodiment, the optimal planning scheme is as follows: Distributed power sources are connected to four nodes: node 8, node 17, node 25, and node 30, with connected capacities of 0.5MW, 0.6MW, 0.4MW, and 0.5MW respectively, for a total connected capacity of 2.0MW.

[0088] The optimized distribution network reduced average network loss by 18.6% (from 0.215MW to 0.175MW) under various typical operating scenarios, reduced average voltage deviation by 32.4%, achieved a renewable energy absorption rate of 96.2%, and met all operational requirements for safety indicators (voltage qualification rate of 99.7% and maximum line load rate of 72%). Compared with existing technologies, the method in this embodiment effectively achieves synergistic optimization of distributed power source access schemes and distribution network operation strategies, demonstrating significant technological advancements.

[0089] Example 2

[0090] The difference between this embodiment and Embodiment 1 is that, for solving the upper planning layer, a genetic algorithm is used instead of the particle swarm algorithm for global search.

[0091] The basic principle of genetic algorithms is to evolve a population through genetic operations such as selection, crossover, and mutation, optimizing and searching for the optimal solution generation by generation. Genetic algorithms are existing technology, and their specific implementation methods are common knowledge in the field.

[0092] The genetic algorithm uses a population size of 150, a maximum number of iterations of 400, a crossover probability of 0.8, and a mutation probability of 0.05. The chromosome encoding method is the same as the particle encoding method in Example 1, employing 16-dimensional encoding (8 binary state variables + 8 continuous capacity variables). The fitness function is the objective function value of the upper planning layer.

[0093] For large-scale distribution network planning problems (more than 100 nodes or more than 20 candidate access points), genetic algorithms have better global search capabilities and can effectively avoid getting trapped in local optima. The remaining steps are the same as in Example 1, and will not be repeated here.

[0094] Example 3

[0095] The difference between this embodiment and Embodiment 1 is that a robust optimization method is used instead of the scenario method to handle uncertainties.

[0096] The basic principle of robust optimization is to replace the probabilistic scenario with an uncertain set and construct a two-layer robust optimization model. The uncertain set is constructed by using historical data of distributed power output and load as a basis to determine the fluctuation range of each uncertain variable (in this embodiment, ±10% of the maximum and minimum values ​​of historical data is taken as the fluctuation range), and representing the uncertainty in the form of an interval.

[0097] When constructing the two-layer robust optimization model, the upper planning layer is the same as in Example 1, while the lower operation layer adopts a robust correspondence form, which ensures the feasibility of operational constraints under the worst uncertain scenarios. The robust optimization model is suitable for scenarios with higher robustness requirements (such as distribution networks with a high proportion of critical loads or power supply areas with strict safety requirements).

[0098] The robust optimization model can be solved using either the Column and Constraint Generation Algorithm (C&CG algorithm) or the Benders decomposition algorithm, both of which are mature algorithms in the existing technology. The basic principle of the C&CG algorithm is to decompose the robust optimization problem into a main problem and subproblems. The main problem solves for the optimal solution under the current worst-case scenario, while the subproblems solve for the dual problem to identify new worst-case scenarios. This process is repeated iteratively until convergence. The remaining steps are the same as in Example 1 and will not be repeated here.

[0099] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A planning and optimization method for integrating distributed generation into a distribution network, characterized in that, Includes the following steps: S1: Obtain the topology parameters of the distribution network, load data of each node, candidate access point information of distributed power sources and their historical output data, and perform data preprocessing. S2: Based on the historical output data of distributed power sources and the load data of each node, a typical operating scenario set is constructed using scenario generation and scenario reduction techniques. The typical operating scenario set is used to characterize the output uncertainty and load uncertainty of distributed power sources. S3: Construct a two-layer optimization model for planning-operation coordinated optimization. The upper layer of the two-layer optimization model is the planning layer, with the goal of minimizing the sum of the annual value cost of distributed power generation investment and construction and the distribution network operation cost. The access location and access capacity of distributed power generation are the upper-layer decision variables. The distribution network operation cost is the expected value of the operation cost under each typical operation scenario weighted by the probability weight of each scenario. The lower layer of the two-layer optimization model is the operation layer, which aims to minimize the sum of the operation cost and the node voltage deviation penalty term under each typical operation scenario, and uses the distribution network operation and scheduling strategy as the decision variable of the lower layer. S4: The two-layer optimization model is iteratively solved using a hybrid solution strategy. The upper planning layer uses the particle swarm algorithm to perform a global search and passes the searched planning scheme to the lower running layer. The lower running layer uses the interior point method to solve the optimal running strategy under the current planning scheme and feeds back the optimal running cost to the upper planning layer. The optimal planning scheme is obtained through iterative coupling between the upper and lower layers. S5: Perform a multi-dimensional comprehensive evaluation of the optimal planning scheme, including economic evaluation, safety evaluation and environmental evaluation, and output the final planning scheme for distributed power source access to the distribution network that meets the evaluation requirements.

2. The planning and optimization method for distributed power generation access to a distribution network according to claim 1, characterized in that: In step S1: the number of nodes in the distribution network, the line impedance parameters between nodes, the line admittance parameters, and the transformer parameters are obtained to form a basic topology database of the distribution network; the annual load curves of each node are obtained according to a preset time period, and are classified according to seasonal type to generate typical daily load data corresponding to each season. The process involves determining the set of candidate access points for distributed power sources, obtaining time-series output data for each candidate access point within a historical period, and normalizing the obtained data.

3. The planning and optimization method for distributed power generation access to a distribution network according to claim 1, characterized in that: In step S2: Based on the historical output data of distributed power sources and the load data of each node, a number of initial uncertainty scenarios containing output fluctuations and load fluctuations of distributed power sources are generated using the Monte Carlo simulation method; the initial uncertainty scenarios are reduced using the fast prior elimination method based on probability distance to obtain a set of typical operating scenarios after the number is reduced; and the occurrence probability of each typical operating scenario is calculated and assigned a corresponding probability weight.

4. The planning and optimization method for distributed power generation access to the distribution network according to claim 1, characterized in that: In step S3: the objective function of the upper planning layer includes the investment and construction cost of distributed power sources in the form of equal annual values, and the network loss cost and the cost of purchasing electricity from the upper-level grid under each typical operating scenario are weighted and summed according to the probability weight of each scenario as the expected value of the distribution network operating cost; the upper planning layer also sets the upper limit constraint of the total access capacity of distributed power sources, the maximum access capacity constraint of each candidate access point, and the upper limit constraint of the penetration rate of distributed power sources.

5. The planning and optimization method for distributed power generation access to a distribution network according to claim 4, characterized in that: In step S3: the upper-level decision variables include binary state variables characterizing whether each candidate access point is connected to a distributed power source, and continuous variables characterizing the distributed power source access capacity at each candidate access point; the upper limit constraint on the total access capacity of distributed power sources means that the sum of the access capacities of all candidate access points shall not exceed the upper limit of the total capacity of distributed power sources allowed to be connected to the distribution network.

6. The planning and optimization method for distributed power generation access to a distribution network according to claim 1, characterized in that: In step S3: The lower-level operation layer constructs a corresponding operation optimization sub-model for each typical operation scenario. The objective function of the operation optimization sub-model is the sum of the operation cost and the node voltage deviation penalty term under each typical operation scenario. The operation cost includes the operation and maintenance cost of distributed power sources and the cost of purchasing electricity from the upper-level grid. The node voltage deviation penalty term is the product of the voltage deviation index of each node in the entire distribution network and the preset penalty weight coefficient. The constraints of the operation optimization sub-model include node power balance constraints, upper and lower limits of voltage amplitude of each node, upper limit of transmission power of each branch, and upper, lower, and upper limits of active power output of distributed power sources.

7. The planning and optimization method for distributed power generation access to a distribution network according to claim 6, characterized in that: In step S3: the lower-level decision variables include the active power output scheduling value of each distributed power source, the reactive power output scheduling value of each distributed power source, the number of switching groups of reactive power compensation devices, and the tap position of the on-load tap-changing transformer; the voltage deviation index of each node of the entire distribution network refers to the sum of the squares of the relative deviations between the actual voltage amplitude of each node and the corresponding node voltage reference value.

8. The planning and optimization method for distributed power generation access to a distribution network according to claim 1, characterized in that: In step S4: Each particle is encoded as a planning scheme containing the access status and access capacity of each candidate access point; the planning scheme obtained in the current iteration is passed to the lower operation layer, and the interior point method is used to solve the operation optimization sub-model for each typical operation scenario to obtain the optimal operation cost and minimum voltage deviation under the current planning scheme; The optimal operating costs of each typical operating scenario obtained from the lower operating layer are weighted by probability and superimposed on the objective function of the upper planning layer to update the fitness value of the current particle, and update the individual optimal position and global optimal position of the particle swarm. It is then determined whether the maximum number of iterations has been reached or whether the global optimal fitness value meets the convergence accuracy. If it does, the global optimal planning scheme is output; otherwise, the iteration continues.

9. The planning and optimization method for distributed power generation access to a distribution network according to claim 8, characterized in that: In step S4: For the non-convex AC power flow constraints in the running optimization sub-model, the running optimization sub-model is first transformed into a convex optimization model using the second-order cone programming relaxation method, and then the original dual interior point method is used to solve the transformed convex optimization model.

10. The planning and optimization method for distributed power generation access to a distribution network according to claim 1, characterized in that: In step S5: Calculate the comprehensive economic index of the optimal planning scheme within a set period. The comprehensive economic index is the sum of the investment cost of distributed power generation, operation and maintenance cost, distribution network loss cost, and cost of purchasing electricity from the upper-level power grid. Calculate the safety indicators of the optimal planning scheme under various typical operating scenarios. The safety indicators include the voltage qualification rate of each node and the maximum load rate of each branch line. Calculate the new energy absorption rate and corresponding carbon emission reduction of the optimal planning scheme as environmental protection indicators; determine whether the economic indicators, safety indicators and environmental protection indicators all meet the preset threshold requirements. If all are met, output the current planning scheme as the final planning scheme. If any indicator does not meet the requirements, return to step S4 to iterate and solve again until the final planning scheme that meets all evaluation indicators is obtained.

Images