Oil field micro-grid two-step optimization scheduling method based on power supply classification

Abstract

The invention provides an oil field micro-grid two-step optimization scheduling method based on power supply classification, and the method comprises the steps: 1, classifying power supplies into a basic load power supply and a frequency modulation power supply; step 2, sampling by adopting a Monte Carlo method; 3, determining a scheduling strategy of a basic load power supply; 4, determining a scheduling strategy based on the BP neural network; and 5, performing simulation calculation to obtain an optimal scheduling scheme of each micro source of the oil field micro-grid. The oilfield microgrid two-step optimization scheduling method based on power supply classification has the advantages of being clear in physical significance, scientific and reasonable in method, high in applicability and reasonable and effective in scheduling strategy.

Description

Technical Field

[0001] The present invention relates to the technical field of oilfield microgrids, and in particular to a two-step optimization scheduling method for oilfield microgrids based on power source classification. Background Art

[0002] Oilfield microgrids achieve power balance between source, grid and load by relying on distributed energy generation and energy storage equipment. However, due to the randomness and intermittent nature of renewable energy output, and considering the working characteristics of the pumping unit, its reverse power generation will affect the dynamic power balance of the system source, grid and load. Therefore, it is of great practical significance to study the safe and economic operation and energy saving and consumption reduction of oilfield microgrids.

[0003] Most of the existing microgrid dispatching methods study deterministic dispatching and formulate dispatching plans without taking into account the uncertainty of renewable energy output. However, in reality, the maximum power that renewable energy can generate is a randomly changing quantity. There is a prediction error between the predicted power and the actual power. When the maximum power that renewable energy can actually generate is less than the power arranged in the dispatching plan, the dispatching plan will be affected by the uncertainty of its output, which will cause a significant increase in operating costs. Therefore, based on the above problems, it is difficult for the existing deterministic optimization dispatching methods to formulate reasonable and effective dispatching plans.

[0004] In the Chinese patent application with application number: CN201310006694.5, a power optimization control method for offshore oilfield power grid is involved. It establishes a large-scale, multi-constraint, nonlinear power optimization mathematical model, takes power generation fuel cost, active network loss, reactive power balance, and voltage residual as optimization targets, and realizes multi-target comprehensive control mode through user control; with flow constraints, voltage constraints, branch transmission power constraints, etc. as constraint conditions, the safety of the system is guaranteed. The interior point method is used for programming on the IntelliJIDEA platform. The invention sets an optimization target suitable for offshore oilfield power grid, realizes coordinated control of active and reactive power of the power grid, realizes optimized automatic control of power grid power, can reduce fuel consumption of offshore oilfield power grid, improve its power generation efficiency, improve the safety and stability of the power grid, and reduce fuel consumption and corresponding pollutant emissions.

[0005] In the Chinese patent application with application number: CN202110389022.1, a power coordination and optimization method for an offshore oilfield group based on shore-based power supply is involved. The power data of the platform equipment in the offshore oilfield platform is obtained, and the power data is sent to the power coordination platform. The power coordination platform receives the power data sent by the offshore oilfield platform, performs line loss analysis and power allocation on the power data, obtains the line loss analysis results and the power allocation results, and sends the results to the remote control platform. The remote control platform is used to obtain the results sent by the power coordination platform and display them for power monitoring personnel to view, and perform remote optimization and adjustment to improve the efficiency of power coordination and optimization of offshore oilfield groups.

[0006] In the Chinese patent application with application number: CN201610172802.X, a method for economic optimization operation and dispatching of microgrids based on an improved quantum genetic algorithm is involved. The microgrid is in a grid-connected mode operation state, including multiple micro sources and loads, the loads include electrical loads and thermal loads, and the micro sources include micro gas turbines, fans, photovoltaic cells, fuel cells, batteries and electric vehicles. The method includes the following steps: S1: obtaining the state information of each load and each micro source in the microgrid; S2: establishing a multi-objective economic dispatch model with the goal of minimizing the operating cost and the pollutant treatment cost; S3: using an improved quantum genetic algorithm to optimize and solve the multi-objective economic dispatch model to obtain the optimal active power of each micro source; S4: controlling the active power output of each micro source according to the optimal active power of each micro source. Compared with the prior art, the invention enables the microgrid composed of distributed power sources to operate in a more economical, flexible and environmentally friendly manner, and can give full play to the power generation advantages of distributed power sources.

[0007] In the Chinese patent application with application number: CN201210516497.3, a short-term power generation prediction method applied to photovoltaic power generation system is involved, which is characterized by comprising two stages: 1. Empirical mode decomposition of the original power generation signal; 2. Constructing a BP neural network prediction model based on genetic algorithm for each decomposition component. The advantages of the present invention are mainly manifested in: accurate photovoltaic power generation prediction in advance can determine the current flow direction of the traditional power grid and the photovoltaic power generation microgrid, formulate the dispatch plan of the power system and reduce the operating cost of the power system; at the same time, energy reserve planning can be formulated in advance to reduce the uncontrollability and intermittency of photovoltaic power generation. The adverse effects on the large power grid, enhance the market competitiveness of photovoltaic power generation. At the same time, the system is also a precedent for applying computer technologies such as artificial neural networks and genetic algorithms to photovoltaic power generation prediction, which has great novelty and practicality.

[0008] The above existing technologies are significantly different from the present invention and fail to solve the technical problem we want to solve. For this reason, we have invented a new two-step optimization scheduling method for oilfield microgrids based on power source classification. Summary of the invention

[0009] The purpose of the present invention is to provide a two-step optimization scheduling method for oilfield microgrids based on power source classification, which has clear physical meaning, takes into account the uncertainty of renewable energy output, has scientific and reasonable methods, strong applicability, and effective scheduling strategy.

[0010] The purpose of the present invention can be achieved by the following technical measures: a two-step optimization scheduling method for oilfield microgrids based on power source classification, the two-step optimization scheduling method for oilfield microgrids based on power source classification comprises:

[0011] Step 1, classify the power sources into basic load power sources and frequency modulation power sources;

[0012] Step 2, using Monte Carlo sampling;

[0013] Step 3, determine the dispatch strategy of the base load power supply;

[0014] Step 4, determine the scheduling strategy based on BP neural network;

[0015] Step 5: Perform simulation calculations to obtain the optimal scheduling plan for each micro-source in the oilfield microgrid.

[0016] The purpose of the present invention can also be achieved by the following technical measures:

[0017] In step 1, the principle of selecting the basic load power source is to select the micro source with relatively small adjustment capacity, slow adjustment speed and relatively weak economy within the adjustment range, and use such micro source as the basic load power source.

[0018] In step 1, the principle of selecting a frequency-modulated power source is whether it has sufficient adjustment capacity, a faster adjustment speed, and whether it has better economy within the adjustment range.

[0019] In step 2, according to the definition of probability, the probability of an event can be estimated by the frequency of occurrence of the event in a large number of tests. When the sample size is large enough, the frequency of occurrence of the event can be considered to be its probability. Therefore, a large number of random samples are first taken from the random variables that affect its reliability, and then these sample values ​​are substituted into the functional formula one by one to determine whether the structure has failed, and finally the failure probability of the structure is obtained.

[0020] In step 2, there are statistically independent random variables Xi (i = 1, 2, 3, ..., k), and their corresponding probability density functions are f(x1), f(x2), ..., f(x k ), the functional formula is Z=g(x1,x2,...,x k ); First, generate N sets of random values ​​x1, x2, ..., x according to the corresponding distribution of each random variablek , calculate the performance function value Z i =g(x1,x2,...,x k )(i=1,2,3...,N), if there are L groups of random numbers corresponding to the function value Z i <0, then when N→∞, according to Bernoulli's law of large numbers and the characteristics of normal random variables, the probability of structural failure can be calculated and the reliability index can be determined, where the probability of structural failure is the number of tests in which structural failure occurs divided by the total number of tests, and the reliability index is the inverse of the probability of structural failure, that is, reliability index = 1 / probability of structural failure.

[0021] In step 3, the particle swarm algorithm is used to optimize the scheduling scheme of the basic load power supply. For a specific particle, that is, a specific scheduling scheme, the power shortage corresponding to a sample under the scheme is obtained. When the power shortage of a sample is known, obtaining the optimal frequency modulation scheme is a linear programming problem. The optimal frequency modulation scheme and its operating cost can be obtained by linear programming. After obtaining the operating cost of the frequency modulation power supply of all samples, the objective function value corresponding to the specific scheduling scheme can be obtained.

[0022] In step 3, the fitness function of a particle is as follows:

[0023]

[0024] In the formula, is the base load power supply objective function, f 2i is the objective function of the frequency modulated power supply under sample i, and N is the number of samplings.

[0025] In step 3, the dispatch strategy of the base load power source is as follows:

[0026] Step 1: Sampling is performed according to the probability distribution to obtain a set of wind turbine and photovoltaic maximum output samples with a sample size of ;

[0027] Step 2: Use the basic load power supply scheduling scheme as the particle swarm particles, initialize the particle swarm particles, obtain the particles that meet the basic load power supply constraints, and initialize the initial velocity of each particle;

[0028] Step 3: The power shortage of each sample in each period is calculated from the power of each particle, and the power shortage of each sample is brought into the linear programming of the frequency modulation power supply to obtain the optimal value of each sample, and then the fitness value of each particle is calculated;

[0029] Step 4: Compare the fitness value of each particle with the historical optimal value, obtain the historical optimal point of each particle and the historical optimal point of the population, and update the speed and position of each particle;

[0030] Step 5: Determine whether each particle still meets the constraints of the basic load power supply. If not, reset the particle. Determine whether each particle coincides with the historical optimal point of the population. If so, reset the particle.

[0031] Step 6: Repeat step 3 to step 5 until the convergence condition is met or the maximum number of iterations is reached.

[0032] In step 4, the BP neural network is a three-layer neural network, including an input layer, a hidden layer and an output layer. The value of the input layer is linearly combined according to the weight and threshold and then calculated by the activation function to obtain the hidden layer data, that is, the value of the jth element of the hidden layer is:

[0033]

[0034] In the formula, x i is the value of the i-th element in the input layer, u j is the value of the jth element in the hidden layer; g1(·) is the activation function from the input layer to the hidden layer; w ij is the weight from the i-th element of the input layer to the j-th element of the hidden layer; b j is the threshold corresponding to the jth element of the hidden layer.

[0035] In step 4, after obtaining the values ​​of each element in the hidden layer, the BP neural network linearly combines the values ​​of each element in the hidden layer according to the weights and thresholds and calculates the activation function to obtain the estimated value of the output layer data at this time, that is, the estimated value of the kth element in the output layer is:

[0036]

[0037] In the formula, y k is the estimated value of the kth element in the output layer; g2(·) is the activation function from the hidden layer to the output layer; L is the number of elements in the hidden layer; w jk is the weight from the jth element in the hidden layer to the kth element in the output layer; b k is the threshold corresponding to the kth element of the output layer.

[0038] In step 4, the estimated value of each element of the output layer is obtained, and the error between the estimated value and the true value of each element of the output layer is obtained. According to the obtained estimated error, a nonlinear optimization algorithm is used to reversely correct each weight and threshold. When the algorithm reaches the convergence condition, the function fitting is considered to be completed.

[0039] In step 4, the premise for the BP neural network to make accurate predictions or formulate better strategies is to use a training set composed of a large amount of historical data for training, perform a large number of samplings based on the probability distribution of the maximum photovoltaic power, and obtain a large number of samples of different power differences in different time periods based on the scheduling plan of the basic load power supply obtained by the two-step scheduling optimization plan and the samples obtained by sampling. Then, linear programming is used to calculate the power generated by each sample frequency modulation power supply in each time period, which is used as the basic data of the training set.

[0040] In step 4, the scheduling strategy based on BP neural network is constructed as follows:

[0041] Step 1: Get the scheduling plan of the basic load power supply;

[0042] Step 2: Perform a large number of sampling according to the probability distribution of the maximum power of wind turbines and photovoltaics to obtain a large number of maximum power samples of wind turbines and photovoltaics;

[0043] Step 3: According to the dispatching plan of the basic load power supply and the power difference of each sample in each period;

[0044] Step 4: Use linear programming to calculate the optimal power allocation for each sample in each time period;

[0045] Step 5: Take the power difference of each sample in the i-th period and the history as the input layer, and the power allocation of each frequency modulation power source in the i-th period as the output layer, build a training set to train the neural network, and obtain the neural network of the power allocation of each frequency modulation power source in the i-th period as the power allocation strategy of the frequency modulation power source in this period;

[0046] Step 6: Repeat step 5 to train the neural network from the 1st period to the 24th period;

[0047] Step 7: If you want to get a better solution, you can bring the obtained scheduling strategy into the two-step scheduling optimization scheme to calculate the power and cost of the frequency modulation power supply, re-optimize the scheduling scheme of the basic load power supply, and retrain the network;

[0048] Step 8: During actual operation, the power difference between this moment and the historical moments is brought into the neural network to obtain the power distribution at this moment.

[0049] In step 5, based on the predicted data of photovoltaic power generation and multi-pumping unit load; distributed power supply capacity, maximum and minimum output, climbing rate and unit power generation cost coefficient; energy storage equipment capacity, charging and discharging power, efficiency and unit power generation cost, a two-step optimization scheduling method is used to obtain the optimal scheduling plan for each micro-source of the oilfield microgrid.

[0050] The two-step optimization scheduling method for oilfield microgrids based on power source classification also includes, after step 5, step 6, establishing microgrid evaluation indicators, the microgrid evaluation indicators are used to evaluate the pros and cons of the scheduling strategy formulated by the microgrid, for this purpose, economic evaluation indicators and renewable energy consumption rates are defined to evaluate the scheduling strategy of the microgrid.

[0051] In step 6, the economic evaluation index and photovoltaic consumption rate index are:

[0052]

[0053]

[0054] In the formula, f is the system operation cost, C i,OM , C i,fule are the operation and maintenance and fuel costs of the micro-source i, is the cost of photovoltaic power generation, C cut is the load shedding cost, C SR is the demand response incentive subsidy; K is the photovoltaic consumption rate, For photovoltaic power output, Predict power for photovoltaics.

[0055] The two-step optimization scheduling method for oilfield microgrid based on power source classification in the present invention adopts the Monte Carlo method to sample the probability density function of renewable energy to obtain a certain number of samples, adopts PSO optimization to find the frequency modulation power supply installation plan and the basic load power supply scheduling plan that minimize the sample cost mean, and uses it as the plan that minimizes the expected cost, and then uses the BP neural network to construct the scheduling strategy of the frequency modulation power supply, which has the advantages of clear physical meaning, scientific and reasonable method, strong applicability, and reasonable and effective scheduling strategy. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a schematic diagram of the structure of an oilfield microgrid in a specific embodiment of the present invention;

[0057] Figure 2 A schematic diagram of a BP neural network in a specific embodiment of the present invention;

[0058] Figure 3 It is a flow chart of a two-step optimization scheduling method in a specific embodiment of the present invention;

[0059] Figure 4 A basic load power source dispatching plan diagram in a specific embodiment of the present invention;

[0060] Figure 5 It is a basic load power source dispatching plan diagram of scheme 2 in a specific embodiment of the present invention;

[0061] Figure 6 A comparison chart of the average daily power generation costs of Scheme 1 and Scheme 3 in a specific embodiment of the present invention;

[0062] Figure 7 A comparison chart of the daily average power generation costs between Scheme 2 and Scheme 4 in a specific embodiment of the present invention;

[0063] Figure 8 A comparison chart of the average daily power generation costs between Scheme 1 and Scheme 5 in a specific embodiment of the present invention;

[0064] Fig. 9 This is a comparison chart of the daily average power generation costs between Scheme 2 and Scheme 6 in a specific embodiment of the present invention.

[0065] Fig.10 The present invention is a flowchart of a specific embodiment of the two-step optimization scheduling method for oilfield microgrid based on power source classification. DETAILED DESCRIPTION

[0066] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0067] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations and / or combinations thereof.

[0068] like Fig.10 As shown, Fig.10 The flowchart of the two-step optimization scheduling method of the oilfield microgrid based on power classification of the present invention. The two-step optimization scheduling method of the oilfield microgrid based on power classification includes:

[0069] 101) Power supply classification

[0070] This method divides power sources into two categories: base load power sources and frequency modulation power sources. The principle of selecting base load power sources is to select those with relatively small adjustment capacity, slow adjustment speed and relatively weak economy within the adjustment range. For example, micro gas turbines and fuel cells have slow adjustment speed, large adjustment cost within the adjustment range and climbing constraints, so these micro sources are used as base load power sources.

[0071] The general principle of selecting a frequency modulation power source is whether it has sufficient adjustment capacity, fast adjustment speed, and whether it has better economy within the adjustment range. For example, batteries have large adjustment capacity, fast adjustment speed, and no adjustment cost problem within the adjustment range, so batteries are generally selected as frequency modulation power sources.

[0072] 102) Monte Carlo sampling

[0073] Monte Carlo sampling principle: According to the definition of probability, the probability of an event can be estimated by the frequency of the event in a large number of tests. When the sample size is large enough, the frequency of the event can be considered to be its probability. Therefore, a large number of random samples can be taken from the random variables that affect its reliability, and then these sample values ​​can be substituted into the functional formula in groups to determine whether the structure fails, and finally the failure probability of the structure can be obtained. The Monte Carlo method is based on this idea for analysis.

[0074] Suppose there are statistically independent random variables Xi (i = 1, 2, 3, ..., k), whose corresponding probability density functions are f(x1), f(x2), ..., f(x k ), the functional formula is Z=g(x1,x2,...,x k ).

[0075] First, according to the corresponding distribution of each random variable, generate N groups of random values ​​x1, x2, ..., x k , calculate the performance function value Z i =g(x1,x2,...,x k )(i=1,2,3...,N), if there are L groups of random numbers corresponding to the function value Z i <0, then when N→∞, according to Bernoulli's law of large numbers and the characteristics of normal random variables, the probability of structural failure can be calculated and the reliability index can be determined, where the probability of structural failure is the number of tests in which structural failure occurs divided by the total number of tests, and the reliability index is the inverse of the probability of structural failure, that is, reliability index = 1 / probability of structural failure.

[0076] From the idea of ​​the Monte Carlo method, it can be seen that this method avoids the mathematical difficulties in structural reliability analysis. Regardless of whether the state function is nonlinear or the random variable is nonnormal, as long as the number of simulations is large enough, a relatively accurate failure probability and reliability index can be obtained.

[0077] 103) Dispatching strategy for base load power supply

[0078] The particle swarm optimization (PSO) algorithm is used to optimize the scheduling scheme of the basic load power supply. For a specific particle, that is, a specific scheduling scheme, the power shortage of a sample under the scheme is obtained. When the power shortage of a sample is known, finding the optimal frequency modulation scheme is a linear programming problem. The optimal frequency modulation scheme and its operating cost can be obtained by linear programming. After obtaining the operating cost of the frequency modulation power supply of all samples, the objective function value corresponding to the specific scheduling scheme can be obtained. The fitness function of a particle is as follows:

[0079]

[0080] In the formula, is the base load power supply objective function, f 2i is the objective function of the frequency modulated power supply under sample i, and N is the number of samplings.

[0081] The dispatch strategy for base load power supply is as follows:

[0082] Step 1: Sampling is performed according to the probability distribution to obtain a set of wind turbine and photovoltaic maximum output samples with a sample size of ;

[0083] Step 2: Use the basic load power supply scheduling scheme as the particle swarm particles, initialize the particle swarm particles, obtain the particles that meet the basic load power supply constraints, and initialize the initial velocity of each particle;

[0084] Step 3: The power shortage of each sample in each period is calculated from the power of each particle, and the power shortage of each sample is brought into the linear programming of the frequency modulation power supply to obtain the optimal value of each sample, and then the fitness value of each particle is calculated;

[0085] Step 4: Compare the fitness value of each particle with the historical optimal value, obtain the historical optimal point of each particle and the historical optimal point of the population, and update the speed and position of each particle;

[0086] Step 5: Determine whether each particle still meets the constraints of the basic load power supply. If not, reset the particle. Determine whether each particle coincides with the historical optimal point of the population. If so, reset the particle.

[0087] Step 6: Repeat step 3 to step 5 until the convergence condition is met or the maximum number of iterations is reached.

[0088] 104) Scheduling strategy based on BP neural network

[0089] Figure 2The figure is a schematic diagram of a BP neural network. The BP neural network is a three-layer neural network, including an input layer, a hidden layer, and an output layer. The values ​​of the input layer are linearly combined according to the weights and thresholds and then calculated by the activation function to obtain the hidden layer data. That is, the value of the jth element of the hidden layer is:

[0090]

[0091] In the formula, u j is the value of the jth element in the hidden layer; g1(·) is the activation function from the input layer to the hidden layer; w ij is the weight from the i-th element of the input layer to the j-th element of the hidden layer; b j is the threshold corresponding to the jth element of the hidden layer.

[0092] After obtaining the values ​​of each element in the hidden layer, the BP neural network linearly combines the values ​​of each element in the hidden layer according to the weights and thresholds and calculates the activation function to obtain the estimated value of the output layer data at this time, that is, the estimated value of the kth element in the output layer is:

[0093]

[0094] In the formula, y k is the estimated value of the kth element in the output layer; g2(·) is the activation function from the hidden layer to the output layer; L is the number of elements in the hidden layer; w jk is the weight from the jth element in the hidden layer to the kth element in the output layer; b k is the threshold corresponding to the kth element of the output layer.

[0095] By obtaining the estimated value of each element in the output layer, we can get the error between the estimated value and the true value of each element in the output layer. According to the estimated error, we use the nonlinear optimization algorithm to reversely correct the weights and thresholds. When the algorithm reaches the convergence condition, the function fitting is considered to be completed.

[0096] The premise for the BP neural network to make accurate predictions or formulate better strategies is to use a training set composed of a large amount of historical data for training, conduct a large number of samplings based on the probability distribution of the maximum photovoltaic power, and obtain a large number of samples of different power differences in different time periods based on the scheduling plan of the basic load power supply obtained by the two-step scheduling optimization plan and the samples obtained by sampling. Then, linear programming is used to calculate the power generated by each sample frequency modulation power supply in each time period, which is used as the basic data of the training set.

[0097] Combining the above analysis, a two-step optimization scheduling method is constructed, and its flow chart is as follows: Figure 3 As shown, the specific solution steps are as follows:

[0098] Step 1: Get the scheduling plan of the basic load power supply;

[0099] Step 2: Perform a large number of sampling according to the probability distribution of the maximum power of wind turbines and photovoltaics to obtain a large number of maximum power samples of wind turbines and photovoltaics;

[0100] Step 3: According to the dispatching plan of the basic load power supply and the power difference of each sample in each period;

[0101] Step 4: Use linear programming to calculate the optimal power allocation for each sample in each time period;

[0102] Step 5: Take the power difference of each sample in the i-th period and the history as the input layer, and the power allocation of each frequency modulation power source in the i-th period as the output layer, build a training set to train the neural network, and obtain the neural network of the power allocation of each frequency modulation power source in the i-th period as the power allocation strategy of the frequency modulation power source in this period;

[0103] Step 6: Repeat step 5 to train the neural network from the 1st period to the 24th period;

[0104] Step 7: If you want to get a better solution, you can bring the obtained scheduling strategy into the two-step scheduling optimization scheme to calculate the power and cost of the frequency modulation power supply, re-optimize the scheduling scheme of the basic load power supply, and retrain the network;

[0105] Step 8: During actual operation, the power difference between this moment and the historical moments is brought into the neural network to obtain the power distribution at this moment.

[0106] 105) Simulation calculation

[0107] Figure 1 This is a schematic diagram of the oilfield microgrid structure, which includes distributed power sources such as photovoltaic power generation, wind power generation, micro gas turbines and fuel cells, energy storage systems and pumping unit loads. The simulation inputs include: predicted data of photovoltaic power generation and multiple pumping unit loads; distributed power capacity, maximum and minimum output,

[0108] Ramp rate and unit power generation cost coefficient; energy storage equipment capacity, charging and discharging power, efficiency and unit power generation cost, etc.; based on the proposed two-step optimization scheduling method, the optimal scheduling plan for each micro-source in the oilfield microgrid is obtained.

[0109] 106) Evaluation indicators

[0110] The microgrid evaluation index is used to evaluate the advantages and disadvantages of the dispatching strategy formulated by the microgrid. To this end, the present invention defines an economic evaluation index and a renewable energy consumption rate to evaluate the dispatching strategy of the microgrid.

[0111] The economic evaluation index and photovoltaic consumption rate index are:

[0112]

[0113]

[0114] In the formula, C eco is the system operating cost, C i,OM , C i,fule are the operation and maintenance and fuel costs of the micro-source i, is the cost of photovoltaic power generation, C cut is the load shedding cost, C SR is the demand response incentive subsidy; K is the photovoltaic consumption rate, For photovoltaic power output, Predict power for photovoltaics.

[0115] The following are several specific embodiments of the present invention.

[0116] Example 1

[0117] In a specific embodiment 1 of the present invention, for Figure 1 The oilfield microgrid with a structure adopts a two-step optimization scheduling strategy. Six schemes are set up under the condition of uncertain renewable energy output to study its impact on the micro-source scheduling strategy: 1) Scheme 1: No pumping unit load optimization is performed, and a two-step optimization scheduling method is adopted, that is, the PSO algorithm is used to optimize the scheduling strategy of the basic load power supply, and the BP neural network is used to formulate the frequency modulation strategy of the frequency modulation power supply; 2) Scheme 2: Under the optimization of the pumping unit load, a two-step optimization scheduling method is adopted, that is, the PSO algorithm is used to optimize the scheduling strategy of the basic load power supply, and the BP neural network is used to formulate the frequency modulation strategy of the frequency modulation power supply; 3) Scheme 3: No pumping unit load optimization is performed, and two 1) Scheme 4: Optimize the load of the pumping unit and adopt a two-step optimization scheduling method, that is, use the PSO algorithm to optimize the scheduling strategy of the basic load power supply, and use the linear programming method to formulate the frequency modulation strategy of the frequency modulation power supply; 2) Scheme 5: Do not optimize the load of the pumping unit, and adopt a one-step optimization scheduling method, that is, directly use the PSO algorithm to optimize the scheduling strategy of the micro-source; 3) Scheme 6: Optimize the load of the pumping unit and adopt a one-step optimization scheduling method, that is, directly use the PSO algorithm to optimize the scheduling strategy of the micro-source.

[0118] The basic load power output results of Scheme 1 and Scheme 2 are as follows: Figure 4 and Figure 5 As shown in the figure, it can be seen that by considering the optimization of the pumping unit load, the load during peak hours is shifted to the off-peak hours, which improves the utilization rate of wind power, alleviates the power supply pressure during peak hours, and improves the economy of the microgrid.

[0119] The BP neural network is used to train and predict the power generated by each frequency modulation power source at each moment. 10,000 samples are obtained by sampling according to the probability distribution of the maximum output of wind turbines and photovoltaics. The daily average electricity price of the frequency modulation strategy using the BP neural network (Scheme 1 and Scheme 2) and the ideal frequency modulation scheme (Scheme 3 and Scheme 4) obtained by linear programming are compared. Figure 6 and 7 By comparison, it can be seen that, in both cases of not considering the optimization of the pumping unit load and considering the optimization of the pumping unit load, the daily average power generation cost of the active power balance according to the frequency regulation strategy of the BP neural network is slightly higher than the frequency regulation scheme under the ideal state, but the cost increase is not large, and the result is relatively ideal.

[0120] According to the probability distribution of the maximum output of wind turbines and photovoltaic power plants, 10,000 samples were obtained. The daily average power generation costs of the one-step scheduling scheme (Scheme 5 and Scheme 6) and the two-step scheduling scheme (Scheme 1 and Scheme 2) were compared without considering the optimization of the pumping unit load and with considering the optimization of the pumping unit load. Figure 8 and 9 As shown in the figure, it can be seen that under the two cases of not considering the optimization of the pumping unit load and considering the optimization of the pumping unit load, the daily average power generation cost obtained by the two-step optimization scheduling method is always lower than the daily average power generation cost obtained by the one-step optimization scheduling method.

[0121] The numerical characteristics of the power generation cost of each scheme are shown in Table 1. It can be seen from the table that regardless of whether the pumping unit load optimization is performed, the proposed two-step optimization scheduling method can effectively reduce the power generation cost and the fluctuation range of the power generation cost when the output of renewable energy is uncertain, which verifies the effectiveness of the constructed two-step scheduling method.

[0122] Table 1 Digital characteristics of power generation costs for each scheme

[0123]

[0124] Example 2

[0125] In the specific embodiment 2 of the application of the present invention, two strategies are set up for comparison and verification of the effectiveness of the present invention. Strategy 1: adopt the two-step planning method proposed in the present invention; Strategy 2: adopt a two-layer optimization method, the upper layer optimizes the load power curve of the pumping unit with the goal of minimizing the fluctuation variance of the load power curve of the pumping unit, and the lower layer formulates the micro-source scheduling strategy with the minimum total cost of the microgrid operation.

[0126] The comparison of the results of strategy one and strategy two is shown in Table 2. The comparison shows that the two-step planning method has better economic benefits than the two-layer optimization method. The main reason is that the upper layer of the hierarchical optimization method only optimizes the oil pumping load, while the two-step optimization method also optimizes the oil pumping load, but the difference is that the latter is optimized on the basis of building a basic load power supply scheduling strategy. This optimization effect is more ideal, so the economic benefits are better. Compared with strategy two, the economy of strategy one is improved by 13.2%, and the energy consumption rate is increased by 6.1%.

[0127] Table 2 Comparison of results between strategy 1 and strategy 2

[0128]

[0129] Example 3

[0130] In the specific embodiment 3 of the application of the present invention, three optimization algorithms are set to compare and verify the effectiveness of the present invention. Algorithm 1: adopts the particle swarm optimization algorithm proposed by the present invention; Algorithm 2: adopts the genetic algorithm; Algorithm 3: adopts the whale optimization algorithm.

[0131] The above three algorithms are used to solve the model respectively, and the average fitness value is taken after 100 tests. The experimental results are shown in Table 3. The particle swarm optimization algorithm has found the global optimal solution when the number of iterations is 20. Compared with the genetic algorithm and the whale optimization algorithm, it has better convergence and faster calculation speed, and is not easy to fall into local extreme values. The experimental results verify the effectiveness of the proposed algorithm.

[0132] Table 3 Comparison of experimental results of different algorithms

[0133]

[0134] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions recorded in the aforementioned embodiments or replace some of the technical features therein with equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0135] Except for the technical features described in the specification, all other technical features are known technologies to those skilled in the art.

Claims

1. A two-step optimization scheduling method for oilfield microgrids based on power source classification, characterized in that: The two-step optimization scheduling method for oilfield microgrid based on power classification includes: Step 1, classify the power sources into basic load power sources and frequency modulation power sources; Step 2, using Monte Carlo sampling; Step 3, determine the dispatch strategy of the base load power supply; Step 4, determine the scheduling strategy based on BP neural network; Step 5: Perform simulation calculations to obtain the optimal scheduling plan for each micro-source in the oilfield microgrid.

2. The two-step optimization scheduling method for oilfield microgrid based on power classification according to claim 1 is characterized in that: In step 1, the principle of selecting the basic load power source is to select the micro source with relatively small adjustment capacity, slow adjustment speed and relatively weak economy within the adjustment range, and use such micro source as the basic load power source.

3. The two-step optimization scheduling method for oilfield microgrid based on power classification according to claim 1 is characterized in that: In step 1, the principle of selecting a frequency-modulated power source is whether it has sufficient adjustment capacity, a faster adjustment speed, and whether it has better economy within the adjustment range.

4. The two-step optimization scheduling method for oilfield microgrid based on power classification according to claim 1 is characterized in that: In step 2, according to the definition of probability, the probability of an event can be estimated by the frequency of occurrence of the event in a large number of tests. When the sample size is large enough, the frequency of occurrence of the event can be considered to be its probability. Therefore, a large number of random samples are first taken from the random variables that affect its reliability, and then these sample values ​​are substituted into the functional formula one by one to determine whether the structure has failed, and finally the failure probability of the structure is obtained.

5. The two-step optimization scheduling method for oilfield microgrid based on power classification according to claim 4 is characterized in that: In step 2, there are statistically independent random variables Xi (i = 1, 2, 3, ..., k), and their corresponding probability density functions are f(x1), f(x2), ..., f(x k ), the functional formula is Z=g(x1,x2,...,x k ); First, generate N sets of random values ​​x1, x2, ..., x according to the corresponding distribution of each random variable k , calculate the performance function value Z i =g(x1,x2,...,x k )(i=1,2,3...,N), if there are L groups of random numbers corresponding to the function value Z i <0, then when N→∞, according to Bernoulli's law of large numbers and the characteristics of normal random variables, the probability of structural failure can be calculated and the reliability index can be determined, where the probability of structural failure is the number of tests in which structural failure occurs divided by the total number of tests, and the reliability index is the inverse of the probability of structural failure, that is, reliability index = 1 / probability of structural failure.

6. The two-step optimization scheduling method for oilfield microgrid based on power classification according to claim 1 is characterized in that: In step 3, the particle swarm algorithm is used to optimize the scheduling scheme of the basic load power supply. For a specific particle, that is, a specific scheduling scheme, the power shortage corresponding to a sample under the scheme is obtained. When the power shortage of a sample is known, obtaining the optimal frequency modulation scheme is a linear programming problem. The optimal frequency modulation scheme and its operating cost can be obtained by linear programming. After obtaining the operating cost of the frequency modulation power supply of all samples, the objective function value corresponding to the specific scheduling scheme can be obtained.

7. The two-step optimization scheduling method for oilfield microgrid based on power classification according to claim 6 is characterized in that: In step 3, the fitness function of a particle is as follows: In the formula, is the base load power supply objective function, f 2i is the objective function of the frequency modulated power supply under sample i, and N is the number of samplings.

8. The two-step optimization scheduling method for oilfield microgrid based on power classification according to claim 7 is characterized in that: In step 3, the dispatch strategy of the base load power source is as follows: Step 1: Sampling is performed according to the probability distribution to obtain a set of wind turbine and photovoltaic maximum output samples with a sample size of ; Step 2: Use the basic load power supply scheduling scheme as the particle swarm particles, initialize the particle swarm particles, obtain the particles that meet the basic load power supply constraints, and initialize the initial velocity of each particle; Step 3: The power shortage of each sample in each period is calculated from the power of each particle, and the power shortage of each sample is brought into the linear programming of the frequency modulation power supply to obtain the optimal value of each sample, and then the fitness value of each particle is calculated; Step 4: Compare the fitness value of each particle with the historical optimal value, obtain the historical optimal point of each particle and the historical optimal point of the population, and update the speed and position of each particle; Step 5: Determine whether each particle still meets the constraints of the basic load power supply. If not, reset the particle. Determine whether each particle coincides with the historical optimal point of the population. If so, reset the particle. Step 6: Repeat step 3 to step 5 until the convergence condition is met or the maximum number of iterations is reached.

9. The two-step optimization scheduling method for oilfield microgrid based on power classification according to claim 1 is characterized in that: In step 4, the BP neural network is a three-layer neural network, including an input layer, a hidden layer and an output layer. The value of the input layer is linearly combined according to the weight and threshold and then calculated by the activation function to obtain the hidden layer data, that is, the value of the jth element of the hidden layer is: In the formula, x i is the value of the i-th element in the input layer, u j is the value of the jth element in the hidden layer; g1(·) is the activation function from the input layer to the hidden layer; w ij is the weight from the i-th element of the input layer to the j-th element of the hidden layer; b j is the threshold corresponding to the jth element of the hidden layer.

10. The two-step optimization scheduling method for oilfield microgrid based on power classification according to claim 9 is characterized in that: In step 4, after obtaining the values ​​of each element in the hidden layer, the BP neural network linearly combines the values ​​of each element in the hidden layer according to the weights and thresholds and calculates the activation function to obtain the estimated value of the output layer data at this time, that is, the estimated value of the kth element in the output layer is: In the formula, y k is the estimated value of the kth element in the output layer; g2() is the activation function from the hidden layer to the output layer; L is the number of elements in the hidden layer; w jk is the weight from the jth element in the hidden layer to the kth element in the output layer; b k is the threshold corresponding to the kth element of the output layer.

11. The two-step optimization scheduling method for oilfield microgrid based on power classification according to claim 10 is characterized in that: In step 4, the estimated value of each element of the output layer is obtained, and the error between the estimated value and the true value of each element of the output layer is obtained. According to the obtained estimated error, a nonlinear optimization algorithm is used to reversely correct each weight and threshold. When the algorithm reaches the convergence condition, the function fitting is considered to be completed.

12. The two-step optimization scheduling method for oilfield microgrid based on power classification according to claim 11 is characterized in that: In step 4, the premise for the BP neural network to make accurate predictions or formulate better strategies is to use a training set composed of a large amount of historical data for training, perform a large number of samplings based on the probability distribution of the maximum photovoltaic power, and obtain a large number of samples of different power differences in different time periods based on the scheduling plan of the basic load power supply obtained by the two-step scheduling optimization plan and the samples obtained by sampling. Then, linear programming is used to calculate the power generated by each sample frequency modulation power supply in each time period, which is used as the basic data of the training set.

13. The two-step optimization scheduling method for oilfield microgrid based on power classification according to claim 12 is characterized in that: In step 4, the scheduling strategy based on BP neural network is constructed as follows: Step 1: Get the scheduling plan of the basic load power supply; Step 2: Perform a large number of sampling according to the probability distribution of the maximum power of wind turbines and photovoltaics to obtain a large number of maximum power samples of wind turbines and photovoltaics; Step 3: According to the dispatching plan of the basic load power supply and the power difference of each sample in each period; Step 4: Use linear programming to calculate the optimal power allocation for each sample in each time period; Step 5: Take the power difference of each sample in the i-th period and the history as the input layer, and the power allocation of each frequency modulation power source in the i-th period as the output layer, build a training set to train the neural network, and obtain the neural network of the power allocation of each frequency modulation power source in the i-th period as the power allocation strategy of the frequency modulation power source in this period; Step 6: Repeat step 5 to train the neural network from the 1st period to the 24th period; Step 7: If you want to get a better solution, you can bring the obtained scheduling strategy into the two-step scheduling optimization scheme to calculate the power and cost of the frequency modulation power supply, re-optimize the scheduling scheme of the basic load power supply, and retrain the network; Step 8: During actual operation, the power difference between this moment and the historical moments is brought into the neural network to obtain the power distribution at this moment.

14. The two-step optimization scheduling method for oilfield microgrid based on power classification according to claim 1 is characterized in that: In step 5, based on the predicted data of photovoltaic power generation and multi-pumping unit load; distributed power supply capacity, maximum and minimum output, climbing rate and unit power generation cost coefficient; energy storage equipment capacity, charging and discharging power, efficiency and unit power generation cost, a two-step optimization scheduling method is used to obtain the optimal scheduling plan for each micro-source of the oilfield microgrid.

15. The two-step optimization scheduling method for oilfield microgrid based on power classification according to claim 1 is characterized in that: The two-step optimization scheduling method for oilfield microgrids based on power source classification also includes, after step 5, step 6, establishing microgrid evaluation indicators, the microgrid evaluation indicators are used to evaluate the pros and cons of the scheduling strategy formulated by the microgrid, for this purpose, economic evaluation indicators and renewable energy consumption rates are defined to evaluate the scheduling strategy of the microgrid.

16. The two-step optimization scheduling method for oilfield microgrid based on power classification according to claim 15 is characterized in that: In step 6, the economic evaluation index and photovoltaic consumption rate index are: In the formula, f is the system operation cost, C i,OM , C i,fule are the operation and maintenance costs and fuel costs of the micro-source i, is the cost of photovoltaic power generation, C cut is the load shedding cost, C SR is the demand response incentive subsidy; K is the photovoltaic consumption rate, For photovoltaic power output, Predict power for photovoltaics.

Images