New energy output uncertainty analysis method and device based on data driving

Abstract

The invention discloses a new energy output uncertainty analysis method and device based on data driving, belongs to the technical field of new energy output analysis, and proposes a data-driven robust optimal power flow model considering wind power and photovoltaic power according to a data-driven mathematical theory on the basis of a deterministic model and an adjustable robust optimal power flow. In the actual operation stage, the optimal solution of the data-driven model shows high robustness, and the uncertainty of renewable energy output can be almost completely immunized. And compared with an adjustable robust optimal model, the data-driven optimal model gives consideration to the economical efficiency of the system on the basis of well ensuring the safety. The method has a good effect in the aspect of continuously improving the permeability caused by the fact that a large number of renewable energy sources are accessed into the electric power system, the uncertainty of the renewable energy sources and the load power can be well solved, and the practicability of a renewable energy source power grid is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of new energy output analysis, and in particular relates to data-driven robust optimization, optimal power flow, renewable energy and second-order cone optimization. Background Art

[0002] For decades, the production of electricity has mostly relied on traditional primary fossil energy, but fossil energy has limited resource reserves, and secondly, the combustion of fossil energy has a great impact on the environment. From the perspective of sustainability and environmental protection, the development and application of renewable energy such as wind power and photovoltaic power generation technology has been accelerated worldwide, and the proportion of renewable energy has increased year by year.

[0003] The access of many new energy sources has brought many problems to the power grid. The most typical one is that the output of RG (wind power generation and photovoltaic power generation system, etc.) is obviously intermittent and volatile, and the load forecast error is large, which makes the line flow and node voltage face the risk of crossing the boundary, and the system security cannot be guaranteed. Therefore, the uncertainty factor should be taken into account in the optimization research of the power system.

[0004] At present, stochastic optimization (SO), fuzzy optimization (FO) and robust optimization (RO) are three powerful methods for solving uncertain optimization problems. Stochastic optimization methods need to fit the probability distribution of uncertain variables through comprehensive analysis of limited sample data, which has certain errors. Fuzzy optimization is to obtain the membership function of parameters with unclear boundaries based on the personal experience of decision makers, which is highly subjective. However, the workload of obtaining these two functions is large. Compared with SO and FO methods, in RO method, the uncertainty of variables is described by a special form of "set" (rather than by probability distribution). When the uncertain variables take any value in the given "set", the system constraints can be satisfied. However, robust optimization sacrifices the optimality of the system under the premise of ensuring the feasibility of the solution. How to improve the economy of the system under the premise of ensuring the safety of the optimization system has become a research hotspot in robust optimization.

[0005] Therefore, a new technical solution is urgently needed in the existing technology to apply this technology to the analysis of new energy output. Summary of the invention

[0006] The technical problem to be solved by the present invention is: to provide a data-driven new energy output uncertainty analysis method and device for analyzing the impact of renewable energy uncertainty on the power grid, and then optimizing the power flow of the power grid connected to the renewable energy; compared with other power flow optimization models in the prior art, the present invention can better solve the problems brought by renewable energy power grids, better balance safety and economy, and be more in line with the actual planning and operation of the power grid.

[0007] Data-driven new energy output uncertainty analysis method, including

[0008] Step 1: Consider renewable energy, establish a nonlinear deterministic optimal power flow model taking renewable energy into account, and use the second-order cone technique to relax the model;

[0009] Step 2: Establish an adjustable robust optimal power flow model taking renewable energy into account, and solve the model to obtain the uncertainty of renewable energy;

[0010] Step 3: Use data-driven methods to robustly optimize the optimal power flow of renewable energy, establish models and solve algorithms;

[0011] Step 4: Use a two-stage simulation calculation process to verify the feasibility of the model and conduct a quantitative analysis of robust optimization in terms of system safety and economy.

[0012] The method for relaxing the model in step 1 is to establish model constraints to relax the model by analyzing the objective function and influencing factors of the model.

[0013] The method for solving the optimal power flow model in step 2 is to establish a set of uncertain variables and then use the duality theory to solve it.

[0014] The uncertainty variable set includes a wind power output uncertainty set,

[0015]

[0016] ||ζ i || ∞ ≤Γ wind , 0≤ζ i ≤1, i∈N wind}

[0017] PV output uncertainty set,

[0018]

[0019] ||τ i || ∞ ≤Γ wind , 0≤τ i ≤1, i∈Nsolar}

[0020] In the formula, Predict active power output for photovoltaics, Predict active power output for wind power; Γ wind , Γ solar Represents the uncertainty level of wind power and photovoltaic output; N wind , N solar represents the wind power access node set and the photovoltaic power station node set, ζ, τ describe the degree of uncertainty, ζ = 0, τ = 0, respectively indicating that the wind power and photovoltaic models are deterministic models.

[0021] A data-driven new energy output uncertainty analysis device, including a modeling unit for establishing a nonlinear deterministic optimal power flow model and an adjustable robust optimal power flow model taking into account renewable energy;

[0022] Model optimization unit, used to solve the model to obtain the uncertainty of renewable energy;

[0023] Data-driven unit, which is used for robust optimization of optimal power flow of renewable energy using data-driven methods, the process of building models and solving algorithms;

[0024] The simulation verification unit is used to verify the feasibility of the model and conduct quantitative analysis on the robust optimization in terms of system safety and economy.

[0025] The model optimization unit includes a wind power output uncertainty set module, which is used to build a wind power output uncertainty set;

[0026] The photovoltaic output uncertainty module is used to establish a photovoltaic output uncertainty set.

[0027] Through the above-mentioned design scheme, the present invention can bring the following beneficial effects: Based on the data-driven new energy output uncertainty analysis method and device, on the basis of the deterministic model and the adjustable robust optimal power flow, according to the data-driven mathematical theory, a data-driven robust optimal power flow model taking into account wind power and photovoltaics is proposed. In the actual operation stage, the optimal solution of the data-driven model shows strong robustness and is almost completely immune to the uncertainty of renewable energy output. And compared with the adjustable robust optimal model, the data-driven optimal model takes into account the economy of the system on the basis of ensuring safety. The present invention has a good effect in coping with the continuous improvement of the penetration rate brought about by the large-scale access of renewable energy to the power system. It can well solve the uncertainty of renewable energy and load power and improve the practicality of renewable energy power grids. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The present invention is further described below with reference to the accompanying drawings and specific embodiments:

[0029] Figure 1 This is a schematic block diagram of the process of the data-driven new energy output uncertainty analysis method of the present invention.

[0030] Figure 2 It is a schematic block diagram of the simulation process of the implementation method of the data-driven new energy output uncertainty analysis method of the present invention. DETAILED DESCRIPTION

[0031] Data-driven new energy output uncertainty analysis method, such as Figure 1 As shown, the following steps are included:

[0032] Step 1: Consider renewable energy, establish a nonlinear deterministic optimal power flow model taking renewable energy into account, and use the second-order cone technique to relax the model;

[0033] The optimal power flow problem is a basic optimization problem in power system analysis. With the development of smart grids, a large number of renewable energy sources are connected to the grid, which will affect the voltage quality, line power flow size and direction, protection coordination, etc.

[0034] From the perspective of economy, the system power generation cost is minimized. However, in the system with new energy access, the power generation of the generator set decreases accordingly with the access of distributed power sources. For the optimal power flow problem that is repeatedly solved, a given decision value can be used as the basic point to calculate the optimal power flow. i It represents the participation factor of each renewable energy group access unit. The power generation of conventional generators will change with the access of distributed power sources, and the change of each power generation unit can be studied at the basic point.

[0035]

[0036] C i (P Gi ) represents the power generation cost of the generator node:

[0037]

[0038] If the average value of the error between the actual value and the predicted value of the system's renewable energy source (RES) is 0, and the covariance matrix is ​​Λ, then the objective function is as follows:

[0039]

[0040] Among them, N G N is the generator set collection; R is the set of renewable energy nodes; σ j Contribute to renewable energy.

[0041] The constraints include the following aspects:

[0042] Let Y ij =G ij +jB ij represents an n×n complex admittance matrix; if the voltage is in polar coordinate form To represent, then the relevant constraints can be expressed in the following form.

[0043] The power equality constraint is:

[0044]

[0045] Where n is the total number of system nodes, is the active output and active load of node i, is the reactive output and reactive load of node i, Predict active and reactive power output for wind power at node i, is the predicted photovoltaic output of node i, G ij ,B ij are the line conductance and susceptance, e and f are the real and imaginary parts of node i.

[0046] The safe operation constraints are:

[0047]

[0048] Since the model is a non-convex linear optimization problem, it is necessary to use the second-order cone relaxation technique to transform it into a convex linear problem. In order to solve this problem, it is necessary to define the variable c ii , c ij ,s ij The relationship between these three variables is: The model after variable transformation is:

[0049]

[0050] Step 2: Establish an adjustable robust optimal power flow model taking renewable energy into account, and solve the model to obtain the uncertainty of renewable energy;

[0051] Commonly used renewable energy sources, such as wind power and photovoltaic power generation, are affected by climate factors, such as wind speed and solar radiation. These influencing factors have strong uncertainty and volatility. Therefore, the uncertainty of renewable energy must be considered when conducting power flow model research.

[0052] Dealing with uncertainty starts from the following aspects.

[0053] (1) Power generation side: Through complementary power generation between power sources, large-scale renewable energy power forecasting, and enhanced reserve capacity, the impact of uncertainty brought by the access of renewable distributed power sources on the power grid can be reduced.

[0054] (2) Grid side: The impact of high-penetration renewable power sources on the grid can be addressed by optimizing the grid structure, enhancing anti-interference capabilities, and strengthening the link between the power generation side and the demand side.

[0055] (3) Demand side: Formulating an appropriate electricity price incentive mechanism for renewable energy can optimize the dispatch of renewable power sources.

[0056] In order to quantitatively consider uncertain factors for robust power flow optimization, it is necessary to consider the decision-making under the worst scenario. First, the set of uncertain variables should be defined.

[0057] Assuming uncertain parameters is the actual active power output of wind power and photovoltaic power, then in, Predict active power output for photovoltaics, is the maximum error of photovoltaic active output, Forecast active power output for wind power, It is the error limit of wind power active output.

[0058] The wind power output uncertainty set can be defined as:

[0059]

[0060] llζilo≤rwind,0≤≤1,i∈Nwind}

[0061] Similarly, the PV output uncertainty set can be defined as:

[0062]

[0063] ||τ i || ∞ ≤Γ wind , 0≤τ i ≤1, i∈N solar}

[0064] By expanding the deterministic optimal power flow model obtained in step 1 and considering the deviation of wind power and photovoltaic output from the predicted values, the adjustable robust optimal power flow model can be written as follows.

[0065] Objective function:

[0066]

[0067] System power constraints:

[0068]

[0069]

[0070] c ij =c ji

[0071] s ij =s ji

[0072] System operation constraints:

[0073]

[0074]

[0075]

[0076]

[0077]

[0078] Second-order cone constraint:

[0079]

[0080] Next, we will solve the model using the duality theory. Assume that ||*|| is the vector R n norm, then the corresponding dual norm can be expressed as ||*|| * . It is defined as follows:

[0081] ||z|| n = sup{z T x||x||≤1}

[0082] So we can get an inequality constraint:

[0083] z T x≤||x||||z||

[0084] Since the above adjustable robust optimal power flow model is linearized into a linear programming problem using the second-order cone technique, in order to clearly express the process of eliminating uncertain parameters, the model can be replaced by its simplified expression, considering the inequality constraints containing uncertain parameters as follows:

[0085] (c T w+d)+(Aw+b) T x≤0

[0086]

[0087] Using the dual norm theory, the robust constraint can be expressed as:

[0088] (b T x+d)+||c+A Tx||q≤0

[0089] Finally, the constraint of the uncertainty parameter is the expression of the line flow constraint:

[0090]

[0091]

[0092] Eliminating the uncertain parameters, we can get the final constraint form:

[0093]

[0094] Step 3: Use data-driven methods to robustly optimize the optimal power flow of renewable energy, establish models and solve algorithms;

[0095] Hypothesis Testing is a method in mathematical statistics to infer the population from the sample based on certain assumptions. Using sample information and a certain probability to make a decision to reject or retain a hypothesis about a population parameter or distribution is called hypothesis testing. Hypothesis testing is to judge whether to accept the null hypothesis and reject the research hypothesis or reject the null hypothesis and accept the research hypothesis based on statistical information. Goodness of fit test is a type of non-parametric test. It is used to test the consistency between the actual observation number and the theoretical number calculated according to a certain hypothesis or model, so as to determine whether the hypothesis or model is consistent with the observation number. The general practice of goodness of fit test is:

[0096] (1) Divide the observations into k categories.

[0097] (2) Obtain n independent observations, the number of observations in the i-th category is O i .

[0098] (3) The probability of the i-th category is p i , The theoretical number of expected numbers of the i-th class is np i =T i .

[0099] (4) O i and T i Compare and judge i and T i Whether the overall degree of non-compliance is caused by chance.

[0100] The principle of uncertain set construction is explained based on typical linear uncertainty optimization problems, and the modeling of single linear constraints is The associated robust constraints are:

[0101]

[0102] Compared to the original prior structural properties, the data S contains more details of the distribution P. By leveraging this additional information, a new set with similar probability guarantees is designed, which is smaller (a subset of the original set) than the traditional set. Therefore, the robust model will obtain a better solution while maintaining its robustness under the new uncertainty set.

[0103] The key to the data-driven framework is to use confidence intervals from hypothesis tests to quantify the information about the distribution P obtained from the data. The construction of the set will depend on the following three points: the prior assumptions about P, ​​the choice of data and hypothesis tests. Different prior assumptions correspond to different hypothesis tests, and different data-driven uncertainty sets will be obtained. Each set has its own geometric properties, computational pressure and modeling level. These sets can contain various characteristics of P, such as skewness, heavy tails and correlation.

[0104] The present invention introduces the concept of divergence to describe two vectors p=(p1, ..., p m ) and q=(q1,…,q m ):

[0105]

[0106] Assume that p in the above formula is the unknown true probability of the uncertain parameter ξ. For the uncertain parameter ξ, there are N historical data, and then the uncertain parameter is divided into m units, ensuring that there are at least 5 or more uncertain parameter historical data in each unit, as shown in the following formula:

[0107]

[0108] From the above formula, the historical data of uncertain parameters can be converted into frequency q = (q1, ...q m ),e T q=1, e is a vector of all 1s and q i Represents the probability of the i-th unit. Its expression is as follows:

[0109]

[0110] For p, we can construct a confidence interval using empirical estimates of q and goodness-of-fit tests. The general goodness-of-fit test is based on the divergence of ψ. Assume that ψ is second-order continuously differentiable around 1 and ψ″(1)>0, then the test statistic roughly obeys one (m-1) degree of freedom Distribution. Based on the test statistic, the 1-α confidence interval of p can be obtained:

[0111] {p∈R m :p≥0,p Te=1,I ψ (p,q)≤ρ}

[0112]

[0113] Assuming that C(S) represents the uncertainty range determined by S, we will next study how to obtain a probability guarantee for region S. First, we need to calculate The minimum value of , p is within the 1-α confidence interval:

[0114]

[0115]

[0116]

[0117] p≥0

[0118] The problem P is a convex optimization problem. According to the historical data of each uncertain parameter, N j Assume that the historical data size represents the jth uncertain parameter, represents the jth uncertain parameter unit i∈V j The corresponding probability represents the jth uncertain parameter unit The uncertain parameters in the aggregation unit (i1, i2, ...i l ) is equal to:

[0119]

[0120] Aggregation unit (i1, i2, ...i l ) is equal to:

[0121]

[0122] Then the problem P can be transformed into the following IP form:

[0123]

[0124]

[0125]

[0126]

[0127]

[0128] At this point, the uncertainty set has been analyzed, and then a robust optimization model with a horizontal line under the data-driven letter can be established. The robust model is as follows:

[0129] maxc T x

[0130]

[0131] Since the error range of wind power output is The photovoltaic output error range is Then the uncertainty parameter U is equivalent to summing the uncertainty set:

[0132]

[0133] The steps of constructing and solving the uncertainty set of the data robust optimal power flow model of the present invention can be expressed by an algorithm, and the algorithm flow is as follows:

[0134] Input: linear optimal power flow model, unit set V, frequency vector q, iteration step τ, confidence interval (1-α), set probability value Output: uncertain set U, robust optimal solution x * , the upper limit of the two-norm ∈

[0135] Step 1: Calculate the given ellipsoidal robust equivalent optimal power flow model and obtain the optimal solution x * .

[0136] Step 2: Calculation

[0137] Step 3: Calculate γ(S, α). If but And the program terminates, otherwise go to step 4.

[0138] Step 4: Let ε = ε + τ and return to step 1.

[0139] Step 4: Use a two-stage simulation calculation process to verify the feasibility of the model and conduct a quantitative analysis of robust optimization in terms of system safety and economy.

[0140] A two-stage simulation calculation process is used to verify the feasibility of the proposed model and analyze the impact of uncertainty on system safety and economy.

[0141] The actual operation / verification phase is equivalent to using the current operation data to verify the robustness and economy of the decisions made in the first phase. That is, the Monte Carlo method is used to simulate and generate n groups of RG active output and node loads, and the security of the optimal decision obtained in the first phase is used. The verification method used is to perform the optimal power flow calculation with the optimal decision quantity as a constant. The convergence of the calculation indicates that the optimal decision is immune to the change of RG output. The high convergence rate indicates that the system has a strong ability to ensure the continued safe operation when the actual RG output is different from the predicted value. Next, the economy of the second phase is compared. When the above optimal power flow calculation does not converge, it means that the optimal decision is unreasonable. At this time, only the power flow is calculated.

[0142] The simulation model can be tested using the IEEE14-node model, in which the uncertainty set is constructed from the historical active output data of wind farms and photovoltaic power stations.

[0143] By using this process for testing, we can find that the uncertainty set obtained by data-driven has better security and better economy.

[0144] The above method flow is a preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. The data-driven new energy output uncertainty analysis method is characterized by: include Step 1: Consider renewable energy, establish a nonlinear deterministic optimal power flow model taking renewable energy into account, and use the second-order cone technique to relax the model; Step 2: Establish an adjustable robust optimal power flow model taking renewable energy into account, and solve the model to obtain the uncertainty of renewable energy; Step 3: Use data-driven methods to robustly optimize the optimal power flow of renewable energy, establish models and solve algorithms; Step 4: Use a two-stage simulation calculation process to verify the feasibility of the model and conduct a quantitative analysis of robust optimization in terms of system safety and economy.

2. The data-driven new energy output uncertainty analysis method according to claim 1 is characterized by: The method for relaxing the model in step 1 is to relax the model by analyzing the objective function and influencing factors of the model and establishing model constraints.

3. The data-driven new energy output uncertainty analysis method according to claim 1 is characterized by: The method for solving the optimal power flow model in step 2 is to establish a set of uncertain variables and then use the duality theory to solve it.

4. The data-driven new energy output uncertainty analysis method according to claim 3 is characterized by: The uncertainty variable set includes a wind power output uncertainty set, PV output uncertainty set, In the formula, Predict active power output for photovoltaics, Predict active power output for wind power; Γ wind , Γ solar Represents the uncertainty level of wind power and photovoltaic output; N wind , N solar represents the wind power access node set and the photovoltaic power station node set, ζ, τ describe the degree of uncertainty, ζ = 0, τ = 0, respectively indicating that the wind power and photovoltaic models are deterministic models.

5. A data-driven new energy output uncertainty analysis device, characterized by: It includes a modeling unit for establishing a nonlinear deterministic optimal power flow model and an adjustable robust optimal power flow model taking into account renewable energy; Model optimization unit, used to solve the model to obtain the uncertainty of renewable energy; Data-driven unit, which is used for robust optimization of optimal power flow of renewable energy using data-driven methods, the process of building models and solving algorithms; The simulation verification unit is used to verify the feasibility of the model and conduct quantitative analysis on the robust optimization in terms of system safety and economy.

6. The data-driven new energy output uncertainty analysis device according to claim 5 is characterized in that: The model optimization unit includes a wind power output uncertainty set module, which is used to build a wind power output uncertainty set; The photovoltaic output uncertainty module is used to establish a photovoltaic output uncertainty set.

Images