A multi-stage scenario set construction method for distribution system planning problems

Abstract

The present invention provides a method for constructing a multi-stage scenario set for a distribution system planning problem, which comprises the following steps: step one, constructing a load growth uncertainty model based on an expected value and a standard deviation of a load growth rate; step two, generating a multi-stage uncertainty scenario based on a Markov chain-Latin hypercube sampling method, and constructing a multi-stage load state scenario tree; step three, performing scenario reduction on the multi-stage load scenario tree based on a reverse sequence reduction method, and generating a multi-stage scenario set for a distribution system; the present invention takes into account the uncertainty of load growth in a distribution system, generates a representative typical scenario set, reduces the error between an actual load scenario and a predicted load scenario, and effectively meets the applicability of distribution system planning problems under load growth uncertainty scenarios.

Description

Technical Field

[0001] The present invention belongs to the field of power system analysis, and in particular relates to a multi-stage scenario set construction method for distribution system planning problems. Background Art

[0002] The distribution system is a highly dynamic and complex network, and its operation is affected by many factors, such as the load change of the distribution system. Its load growth is affected by many factors such as geographical space, economic level, and living habits, and has a strong uncertainty characteristic. The uncertainty of load growth also brings great difficulties to the economic operation of the power grid. In the traditional distribution system planning method, there is an error between the actual load scenario and the predicted load scenario. The planning scheme determined in the predicted scenario often does not work well in the actual scenario. Therefore, how to take into account the uncertainty of load growth and select a representative set of typical operation scenarios has become an important research content.

[0003] Based on this background, the present invention proposes a multi-stage scenario set construction method for the distribution system planning problem. By generating a load scenario set that is more in line with actual needs, the rationality and feasibility of the planning scheme under the multi-stage load growth uncertainty scenario are improved, and the applicability of the distribution system under the load growth uncertainty scenario is effectively met. Summary of the invention

[0004] The purpose of the present invention is to provide a multi-stage scenario set construction method for distribution system planning problems, which is used to solve the technical problem of multi-stage scenario set generation that ignores the uncertainty of load growth in the prior art.

[0005] In order to achieve the above technical objectives, the present invention adopts the following scheme:

[0006] A multi-stage scenario set construction method for distribution system planning problem, the method comprising the following steps:

[0007] S1 constructs a load growth uncertainty model based on the expected value and standard deviation of the load growth rate, obtains the estimated value of the future planned load growth rate, and calculates the load peak value;

[0008] The expected value of the load growth rate is determined in two ways according to whether the influencing variable has a planned value. The standard deviation of the load growth rate is approximately estimated using the standard deviation of the historical load growth rate.

[0009] S2 generates multi-stage uncertainty scenarios based on the Markov Chain-Latin hypercube sampling (MC-LHS) method, constructs a multi-stage load state scenario tree, and converts the scenario tree into a scenario trajectory;

[0010] The horizontal axis of the scenario tree is the prediction domain. The entire prediction domain is divided into several stages. Each stage contains multiple dimensional states. Each dimensional state is composed of several nodes. Each node is a child node of the corresponding node in the previous dimensional state. The probability of each node is the sum of the probabilities of all its child nodes.

[0011] The path from the root node to the leaf node is called a scene track. All possible scene tracks together constitute the entire scene tree. The number of scene tracks is equal to the number of leaf nodes in the last stage of the scene tree.

[0012] S3 performs scenario reduction on the multi-stage load scenario tree obtained in step S2 based on the reverse sequence reduction method to generate a multi-stage scenario set for the distribution system;

[0013] For all scenario trajectories of the multi-stage load state scenario tree, starting from the final stage node, the load state sequence is gradually generated in reverse order, and the probability distances of different state sequences are reduced to retain the typical load state sequences and ensure that the retained state sequences still obey the probability distribution of the state sequences before reduction.

[0014] In S1 of the present invention, when determining the expected value of the load growth rate, if there is a future planning value, combined with the time value of the current year, the average growth rate of future years is calculated and used as the mean of the future random change rate. If there is no future planning value, the mean of the historical change rate can be used as the mean of the future change rate.

[0015] In S1 of the present invention, an uncertainty model of the load growth rate is established. The specific method is as follows: the random change of the load growth rate is affected by multiple independent factors that cannot produce a dominant effect. Therefore, it can be considered that the load growth rate approximately obeys the normal distribution. The following formula is used to represent the random distribution of the load growth rate:

[0016]

[0017] Among them, λ Load is the estimated value of future planned load growth rate, μ Load , σ Load are the expected value and standard deviation of load growth rate respectively.

[0018] In S1 of the present invention, the load peak is calculated according to the estimated value of the future planned load growth rate, and the specific formula is:

[0019] P Load =P B (1+λ Load )

[0020] Among them, P Load is the load peak, P B is the base load peak.

[0021] In S2 of the present invention, when the scene is generated according to the MC-LHS method, the specific process is as follows:

[0022] S21 sets the number of stages to k, the total number of stages to N, the number of load samples generated to S, the number of scenarios retained between stages to M, sets k=0, and enters S22;

[0023] S22 Input load expectation set E of stage k (k) , standard deviation set Q (k) (Q (k) =0.1·E (k) ), let the initial state r of stage k be (k) =[E (k) ,Q (k) ], based on the probability distribution function of load growth, the initial state r is calculated by LHS (k) Sample S load samples and construct state groups Update the transition probability of each state according to the Markov chain and construct a transition probability group Enter S23;

[0024] If S>M in S23, use the synchronous back-substitution reduction method to reduce the scene and enter S24;

[0025] S24 Update stage k state group Update transition probability group Enter S25;

[0026] S25 determines whether the number of scenes after reduction is less than or equal to the number of scenes to be retained. If S≤M, proceed to S26; otherwise, return to S23;

[0027] S26 Order is the actual load growth rate at stage k;

[0028] S27 sets k=k+1 and updates the load expectation of stage k Load standard deviation Update the initial state r (k) =[E (k) ,Q (k) ], enter S28;

[0029] If k≤N in S28, return to S22, otherwise, enter S29;

[0030] S29 completes multi-stage scene generation and outputs multi-stage state groups Multi-stage transition probability group

[0031] Among them, LHS is used to generate load samples. The specific method is as follows:

[0032] (1) Sampling, defining the initial sample The cumulative distribution function of the initial sample is Will The value interval of is divided into s sample spaces, the occurrence probability of each sample space is 1 / s, the sampling method is median sampling, and the sample value is obtained by inverse transformation using the mapping function According to the above steps, a sample matrix W with r rows and s columns can be generated. α ;

[0033] (2) Sorting: The sample values ​​are independent of each other. The correlation of the sample values ​​is eliminated by rearranging them. The samples are rearranged based on the Cholesky decomposition method to obtain the processed sample matrix W. β ;

[0034] In S3 of the present invention, a reverse sequence reduction method is used to reduce scenes, and the specific process is as follows:

[0035] S31 sets i=1, and u=u+N in the initial stage;

[0036] S32 Update phase u state sequence The corresponding probability Stage u initial scene set Ω (u) , set It is an empty set, the number of iterations k=1, and the process goes to S33;

[0037] S33 calculates the scene that needs to be cut at the kth iteration Make the following formula take the minimum value;

[0038]

[0039] in, is the set at the k-1th iteration of stage u

[0040] S34 Delete scene from D Will Add to collection

[0041] S35 will have its scenes cut The probability of is added to the scenario closest to its probability;

[0042] S36 repeats steps S33 to S35 until Ω (u) The number of scenes in N s Equal to the number of scenes N finally retained c , enter S37;

[0043] S37 sets i=i+1, updates the number of stages to u=u+N-i+1, and proceeds to S38;

[0044] S38: If u+N-i+1>u+T, return to S32; otherwise, enter S39;

[0045] S39 outputs the reserved status sequence The corresponding probability

[0046] The probability distance between the state sequences is specifically calculated as follows:

[0047]

[0048] Among them, the i-th group sequence is recorded as The jth group of sequences is recorded as in and Have a common parent node

[0049] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0050] 1. The present invention aims at a multi-stage and multi-dimensional scenario set, takes into account the uncertainty of load growth in the distribution system, and proposes a scenario generation method suitable for distribution network planning. The simulation of typical operation scenarios of the distribution system is more accurate and more universal;

[0051] 2. The present invention reduces the error between the actual load scenario and the predicted load scenario, and the method enhances the applicability of the distribution system planning scheme under the scenario of load growth uncertainty. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a schematic flow chart of the method of the present invention. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] Figure 1 It is a schematic flow chart of the method of the present invention, which introduces the basic steps of the method of the present invention.

[0055] like Figure 1 A multi-stage scenario set construction method for a distribution system planning problem is shown, which specifically includes the following steps:

[0056] S1 constructs a load growth uncertainty model based on the expected value and standard deviation of the load growth rate, obtains the estimated value of the future planned load growth rate, and calculates the load peak value;

[0057] The expected value of the load growth rate is determined in two ways according to whether the influencing variable has a planned value. The standard deviation of the load growth rate is approximately estimated using the standard deviation of the historical load growth rate.

[0058] S2 generates multi-stage uncertainty scenarios based on the Markov Chain-Latin hypercube sampling (MC-LHS) method, constructs a multi-stage load state scenario tree, and converts the scenario tree into a scenario trajectory;

[0059] The horizontal axis of the scenario tree is the prediction domain. The entire prediction domain is divided into several stages. Each stage contains multiple dimensional states. Each dimensional state is composed of several nodes. Each node is a child node of the corresponding node in the previous dimensional state. The probability of each node is the sum of the probabilities of all its child nodes.

[0060] The path from the root node to the leaf node is called a scene track. All possible scene tracks together constitute the entire scene tree. The number of scene tracks is equal to the number of leaf nodes in the last stage of the scene tree.

[0061] S3 performs scenario reduction on the multi-stage load scenario tree obtained in step 2 based on the reverse sequence reduction method to generate a multi-stage scenario set for the distribution system;

[0062] For all scenario trajectories of the multi-stage load state scenario tree, starting from the final stage node, the load state sequence is gradually generated in reverse order, and the probability distances of different state sequences are reduced to retain the typical load state sequences and ensure that the retained state sequences still obey the probability distribution of the state sequences before reduction.

[0063] In this embodiment, when the expected value of the load growth rate is determined in the above step S1, if there is a future planning value, combined with the time value of the current year, the average growth rate of future years is calculated and used as the mean of the future random change rate. If there is no future planning value, the mean of the historical change rate can be used as the mean of the future change rate.

[0064] In this embodiment, the above step S1 establishes a load growth rate uncertainty model, and the specific method is: the random change of the load growth rate is affected by multiple independent and non-dominant factors, so it can be considered that the load growth rate approximately obeys the normal distribution, and the following formula is used to represent the random distribution of the load growth rate:

[0065]

[0066] Among them, λ Load is the estimated value of future planned load growth rate, μ Load , σ Load are respectively the expected value and standard deviation of the load growth rate. In this embodiment, the expected value of the load growth rate is set to 30%, and the variance value is set to 10% of the expected value.

[0067] In this embodiment, the above step S1 calculates the load peak value according to the estimated value of the future planned load growth rate, and the specific formula is:

[0068] P Load =P B (1+λ Load )

[0069] Among them, P Load is the load peak, P B is the base load peak.

[0070] In this embodiment, when the above step S2 generates a scene according to the MC-LHS method, the specific process is as follows:

[0071] S21 sets the number of stages to k, the total number of stages to N, the number of load sample generation to S, the number of scenarios reserved between stages to M, and k=0, and then goes to step S22. In this embodiment, the total number of stages N is set to 3, and the number of scenarios reserved between stages M is set to 9;

[0072] S22 Input load expectation set E of stage k (k) , standard deviation set Q (k) (Q (k) =0.1·E (k) ), let the initial state r of stage k be (k) =[E (k) ,Q (k) ], based on the probability distribution function of load growth, the initial state r is calculated by LHS (k) Sample S load samples and construct state groups Update the transition probability of each state according to the Markov chain and construct a transition probability group Go to step S23;

[0073] If S>M in step S23, use the synchronous back-substitution reduction method to reduce the scene and go to step S24;

[0074] S24 Update stage k state group Update transition probability group Go to step S25;

[0075] S26 determines whether the number of scenes after reduction is less than or equal to the number of scenes to be retained. If S≤M, proceed to step S26; otherwise, return to step S23;

[0076] S26 Order is the actual load growth rate at stage k;

[0077] S27 sets k=k+1 and updates the load expectation of stage k Load standard deviation Update the initial state r (k) =[E (k) ,Q (k) ], proceed to step S28;

[0078] S28 If k≤N, return to step S22, otherwise, go to step S29;

[0079] S29 completes multi-stage scene generation and outputs multi-stage state groups Multi-stage transition probability group

[0080] The specific method of using LHS to generate load samples is as follows:

[0081] (1) Sampling, defining the initial sample The cumulative distribution function of the initial sample is Will The value interval of is divided into s sample spaces, the occurrence probability of each sample space is 1 / s, the sampling method is median sampling, and the sample value is obtained by inverse transformation using the mapping function According to the above steps, a sample matrix W with r rows and s columns can be generated. α ;

[0082] (2) Sorting: The sample values ​​are independent of each other. The correlation of the sample values ​​is eliminated by rearranging them. The samples are rearranged based on the Cholesky decomposition method to obtain the processed sample matrix W. β ;

[0083] In this embodiment, the above step S3 uses the reverse sequence reduction method to reduce scenes, and the specific process is as follows:

[0084] S31 sets i=1, and u=u+N in the initial stage;

[0085] S32 Update phase u state sequence The corresponding probability Stage u initial scene set Ω (u) , set is an empty set, the number of iterations k=1, and the process goes to step S33;

[0086] S33 calculates the scene that needs to be cut at the kth iteration Make the following formula take the minimum value;

[0087]

[0088] in, is the set at the k-1th iteration of stage u

[0089] S34 Delete scene from D Will Add to collection

[0090] S35 will have its scenes cut The probability of is added to the scenario closest to its probability;

[0091] S36 repeats steps S33 to S35 until Ω (u) The number of scenes in N s Equal to the number of scenes N finally retained c , proceed to step S37;

[0092] S37 sets i=i+1, updates the number of stages to u=u+N-i+1, and proceeds to step S38;

[0093] S38 If u+N-i+1>u+T, return to step 2), otherwise, go to step S39;

[0094] S39 outputs the reserved status sequence The corresponding probability

[0095] The specific calculation formula of the probability distance between the state sequences is as follows:

[0096]

[0097] Among them, the i-th group sequence is recorded as The jth group of sequences is recorded as in and Have a common parent node

[0098] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

[0099] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is only limited by the claims and their full scope and equivalents.

[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and are not intended to limit the present invention. A person skilled in the art should understand that the specific implementation of the present invention can be modified or replaced by equivalents, but these modifications or changes are within the scope of protection of the pending claims.

Claims

1. A multi-stage scenario set construction method for distribution system planning problem, characterized in that: The method comprises the following steps: S1 constructs a load growth uncertainty model based on the expected value and standard deviation of the load growth rate, obtains the estimated value of the future planned load growth rate, and calculates the load peak value; The expected value of the load growth rate is determined in two ways according to whether the influencing variable has a planned value. The standard deviation of the load growth rate is approximately estimated by the standard deviation of the historical load growth rate. S2 generates multi-stage uncertainty scenarios based on the Markov Chain-Latin hypercube sampling (MC-LHS) method, constructs a multi-stage load state scenario tree, and converts the scenario tree into a scenario trajectory; The horizontal axis of the scenario tree is the prediction domain. The entire prediction domain is divided into several stages. Each stage contains multiple dimensional states. Each dimensional state is composed of several nodes. Each node is a child node of the corresponding node in the previous dimensional state. The probability of each node is the sum of the probabilities of all its child nodes. The path from the root node to the leaf node is called a scene track. All possible scene tracks together constitute the entire scene tree. The number of scene tracks is equal to the number of leaf nodes in the last stage of the scene tree. S3 performs scenario reduction on the multi-stage load scenario tree obtained in step S2 based on the reverse sequence reduction method to generate a multi-stage scenario set for the distribution system; For all scenario trajectories of the multi-stage load state scenario tree, starting from the final stage node, the load state sequence is gradually generated in reverse order, and the probability distances of different state sequences are reduced to retain the typical load state sequences and ensure that the retained state sequences still obey the probability distribution of the state sequences before reduction.

2. A multi-stage scenario set construction method for distribution system planning problems according to claim 1, characterized in that: Determine the expected value of the load growth rate by: If there are future planning values, combined with the time value of the current year, the average annual growth rate in the future can be calculated and used as the mean of the future random change rate. If there are no future planning values, the mean of the historical change rates can be used as the mean of the future change rate.

3. The multi-stage scenario set construction method for distribution system planning problem according to claim 1 is characterized in that: The load growth uncertainty model is established in the following specific method: The random change of load growth rate is affected by multiple independent factors that cannot produce dominant factors. Therefore, it can be considered that the load growth rate approximately obeys the normal distribution. The following formula is used to express the random distribution of load growth rate: Among them, λ Load is the estimated value of future planned load growth rate, μ Load , σ Load are the expected value and standard deviation of load growth rate respectively.

4. The method for constructing a multi-stage scenario set for a power distribution system planning problem according to claim 1, characterized in that: The load peak is calculated based on the estimated value of the future planned load growth rate. The specific formula is: P Load =P B (1+λ Load ) Among them, P Load is the load peak, P B is the base load peak.

5. The multi-stage scenario set construction method for distribution system planning problem according to claim 1, characterized in that: When the scene is generated according to the MC-LHS method, the specific process is as follows: S21 sets the number of stages to k, the total number of stages to N, the number of load samples generated to S, the number of scenarios retained between stages to M, sets k=0, and enters S22; S22 Input load expectation set E of stage k (k) , standard deviation set Q (k) (Q (k) =0.1·E (k) ), let the initial state r of stage k be (k) =[E (k) ,Q (k) ], based on the probability distribution function of load growth, the initial state r is calculated by LHS (k) Sample S load samples and construct state groups Update the transition probability of each state according to the Markov chain and construct a transition probability group Enter S23; If S>M in S23, use the synchronous back-substitution reduction method to reduce the scene and enter S24; S24 Update stage k state group Update transition probability group Enter S25; S25 determines whether the number of scenes after reduction is less than or equal to the number of scenes to be retained. If S≤M, proceed to S26; otherwise, return to S23; S26 Order is the actual load growth rate at stage k; S27 sets k=k+1 and updates the load expectation of stage k Load standard deviation Update the initial state r (k) =[E (k) ,Q (k) ], enter S28; If k≤N in S28, return to S22, otherwise, enter S29; S29 completes multi-stage scene generation and outputs multi-stage state groups Multi-stage transition probability group 6. A multi-stage scenario set construction method for distribution system planning problems according to claim 5, characterized in that: Use LHS to generate load samples. The specific method is as follows: (1) Sampling, defining the initial sample The cumulative distribution function of the initial sample is Will The value interval of is divided into s sample spaces, the occurrence probability of each sample space is 1 / s, the sampling method is median sampling, and the sample value is obtained by inverse transformation using the mapping function According to the above steps, a sample matrix W with r rows and s columns can be generated. α ; (2) Sorting: The sample values ​​are independent of each other. The correlation of the sample values ​​is eliminated by rearranging them. The samples are rearranged based on the Cholesky decomposition method to obtain the processed sample matrix W. β .

7. A multi-stage scenario set construction method for distribution system planning problem according to claim 1, characterized in that: In S3, the reverse sequence reduction method is used to reduce scenes. The specific process is as follows: S31 sets i=1, and u=u+N in the initial stage; S32 Update phase u state sequence The corresponding probability Stage u initial scene set Ω (u) , set It is an empty set, the number of iterations k=1, and the process goes to S33; S33 calculates the scene that needs to be cut at the kth iteration Make the following formula take the minimum value; in, is the set at the k-1th iteration of stage u S34 Delete scene from D Will Add to collection S35 will have its scenes cut The probability of is added to the scenario closest to its probability; S36 repeats steps S33 to S35 until Ω (u) The number of scenes in N s Equal to the number of scenes N finally retained c , enter S37; S37 sets i=i+1, updates the number of stages to u=u+N-i+1, and proceeds to S38; S38: If u+N-i+1>u+T, return to step 2), otherwise, go to S39; S39 outputs the reserved status sequence The corresponding probability 8. A multi-stage scenario set construction method for distribution system planning problem according to claim 7, characterized in that: The probability distance between state sequences is calculated as follows: Among them, the i-th group sequence is recorded as The jth group of sequences is recorded as in and Have a common parent node

Images