Typical day design boundary extraction method for integrated energy system based on optimization

Abstract

The present invention relates to a typical day extraction method (K-MIMLP) based on optimization, which comprehensively considers the key factors affecting the planning results in the design boundary, including extremeness, cumulativeness and morphology, so that the accuracy of the optimization results is much higher than the typical day extraction method commonly used at present. It mainly includes the following steps: 1) Collecting 8760h of original time series, including cooling load, heating load, electric load, renewable energy output, etc.; 2) Standardizing the data, this patent adopts z-score standardization; 3) Eliminating extreme days (each day with an attribute reaching the annual peak is considered to be an extreme day); 4) Setting the number of typical days, target weights, and constraint parameters in sequence; 5) Adding the extreme days that have been eliminated again after reaching convergence; 6) Outputting the typical day weight; 7) Using peak index, cumulative index, and morphological index to evaluate its accuracy.

Description

Technical Field

[0001] The present invention relates to the field of integrated energy system planning and design, and in particular to an optimization-based integrated energy system typical day design boundary extraction method. Background Art

[0002] The integrated energy system utilizes multiple energy sources, breaks down the barriers between different energy supply systems, and realizes the overall coordination, cooperation and optimization of multiple subsystems in various stages such as planning, design, construction and operation. The key factor in its application is to establish an economical and reliable planning scheme to determine the optimal capacity design of the integrated energy system and ensure the balance between supply and consumption. At present, the effective method to obtain the design scheme of the integrated energy system is to establish an optimization model for optimization.

[0003] The premise of establishing an optimization model is to determine the design boundary. The design boundary not only determines the problem of equipment type and capacity matching, but also determines the energy balance during operation. It is the premise and basis for ensuring the reliability and rationality of the design scheme. Therefore, adopting a set of typical day design boundaries is the premise and basis for ensuring the accuracy of the optimization model while improving the solution efficiency. However, the current typical day selection method ignores the purpose of serving the planning and design of the integrated energy system, and the consideration of key information is not comprehensive enough. The typical day is determined only from a statistical perspective, which leads to poor accuracy of the design scheme optimization results. Therefore, the present invention proposes a typical day design boundary selection method that retains the "key information" of the original design boundary / time series as much as possible, in order to overcome the defects of the current method and provide more reasonable guidance for the planning and design of the integrated energy system. Summary of the invention

[0004] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a typical day design boundary extraction method for integrated energy system planning and design.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] A typical day design boundary extraction method for an integrated energy system based on optimization, characterized by comprising the following steps:

[0007] S1. Collect 8760 hours of original time series, including cooling load, heating load, electric load, renewable energy output, etc.;

[0008] S2. Standardize the data. This patent uses z-score standardization;

[0009] S3, eliminate extreme days (a day when each attribute reaches the annual peak is considered an extreme day);

[0010] S4. Set the typical days, target weight, and constraint parameters in sequence;

[0011] S5, after reaching convergence, add the removed extreme days again;

[0012] Furthermore, the step S5 reaches convergence when the objective function reaches a minimum value, and the sum of the dissimilarity between the Euclidean distance and the Pearson correlation coefficient is minimized as the optimization goal, and constraints are imposed from three aspects: extreme value, cumulative value, and morphology.

[0013] Furthermore, the objective function expression of step S5 is as follows:

[0014]

[0015] Among them, bin i,j is a binary variable. If the jth day is replaced by the ith day, bin i，j =1, otherwise bin i，j =0.d i，j represents the Euclidean distance between the i-th day and the j-th day. i,j represents the Pearson correlation coefficient between day i and day j, 1-r i,j Then represents the difference between the two, cov represents the covariance, and σ represents the standard deviation. eu 、V pc is the corresponding target weight. σ is a dimensionless coefficient.

[0016] Furthermore, the constraint condition expression of step S5 is as follows:

[0017]

[0018]

[0019] Formula (4) forces each day j to be replaced by only one typical day i. Formula (5) is used to calculate the weight W of the typical day i In formulas (6) to (7), is also a binary variable. If the i-th day is selected as a typical day, otherwise K is the number of typical days, which can be set by the user. Formulas (8) to (13) constrain the deviation between the extracted typical days and the original time series in terms of extreme values, cumulative values, and forms.

[0020] S6. Output typical day weight

[0021] S7. Use peak index, cumulative index and morphological index to evaluate its accuracy.

[0022] Compared with the prior art, the present invention has the following advantages:

[0023] First, the proposed K-MIMLP can effectively extract the key information of the original time series. In the proposed K-MIMLP model, relevant indicators such as extreme values, cumulative values, and forms of the original data are added as constraints, which can flexibly adjust the weights of the extreme scenarios, annual cumulative values, and forms of the selected typical days, making the planning scheme more reasonable.

[0024] Second, K-MIMLP can effectively reduce the computational complexity of planning problems without sacrificing the accuracy of the decision maker's design. This will help further advance the in-depth study of integrated energy system planning and design problems in terms of multiple location scales, multiple time scales, and input data uncertainty. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is the hourly heat load of the scientific research station throughout the year.

[0026] Figure 2 It is the hourly electricity load of the scientific research station throughout the year.

[0027] Figure 3 This is a schematic diagram of the scientific research station’s energy system.

[0028] Figure 4 is the typical daily load selected using the AM method.

[0029] Figure 5 is the typical daily load selected using the K-means method.

[0030] Figure 6 is the typical daily load selected using the K-MIMLP method.

[0031] Figure 7 These are the evaluation index results for different typical days.

[0032] Figure 8 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0033] The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0034] Taking the hourly thermal power load of a scientific research station in Antarctica as an example, this method is applied to the selection of its typical daily load. The specific steps are as follows:

[0035] S1. Collect 8760 hours of original time series, including cooling load, heating load, electricity load, renewable energy output, etc. The hourly thermal and electricity load of the scientific research station throughout the year is as follows: Figure 1-2The energy system diagram is shown in Figure 3 As shown, the optional equipment models and equipment parameters are shown in Table 1;

[0036]

[0037] Table 1

[0038] S2. Standardize the data. This patent uses z-score standardization;

[0039] S3, eliminate extreme days (a day when each attribute reaches the annual peak is considered an extreme day);

[0040] S4. Set the typical days, target weight, and constraint parameters in sequence;

[0041] S5, after reaching convergence, add the removed extreme days again;

[0042] Furthermore, the step S5 reaches convergence when the objective function reaches a minimum value, and the sum of the dissimilarity between the Euclidean distance and the Pearson correlation coefficient is minimized as the optimization goal, and constraints are imposed from three aspects: extreme value, cumulative value, and morphology.

[0043] Furthermore, the objective function expression of step S5 is as follows:

[0044]

[0045] Among them, bin i,j is a binary variable. If the jth day is replaced by the ith day, bin i,j =1, otherwise bin i，j =0.d i，j represents the Euclidean distance between the i-th day and the j-th day. i,j represents the Pearson correlation coefficient between day i and day j, 1-r i,j Then represents the difference between the two, cov represents the covariance, and σ represents the standard deviation. eu 、V pc is the corresponding target weight. σ is a dimensionless coefficient.

[0046] Furthermore, the constraint condition expression of step S5 is as follows:

[0047]

[0048]

[0049] Formula (4) forces each day j to be replaced by only one typical day i. Formula (5) is used to calculate the weight W of the typical day i In formulas (6) to (7), is also a binary variable. If the i-th day is selected as a typical day, otherwise K is the number of typical days, which can be set by the user. Formulas (8) to (13) constrain the deviation between the extracted typical days and the original time series in terms of extreme values, cumulative values, and forms.

[0050] S6. Output typical day weights. The typical day loads selected by AM, K-means, and K-MIMLP methods are as follows: Figure 4-6 As shown in Table 2, the optimal configuration results obtained by different typical day selection methods are shown in Table 2;

[0051]

[0052]

[0053] Table 2

[0054] according to Figure 4-6 It can be clearly seen that AM completely loses the extreme scenario, while K-means and K-MIMLP retain the extreme scenario to the maximum extent, especially K-MIMLP. This shows that the most unfavorable operating condition constraint added by K-MIMLP effectively improves the retention of the most unfavorable operating condition points on both sides of the source and load on a typical day.

[0055] For the optimization configuration results, (1) Renewable energy production equipment: All three schemes underestimated the energy supply potential of renewable energy. In general, the deviation was large, but K-MIMLP had the smallest deviation. (2) Conventional energy conversion equipment: AM and K-means had the same configuration results. The total heating capacity of K-MIMLP was 6.1% lower, and the total power supply capacity was consistent with the reference case, i.e., 0.0%. K-MIMLP has outstanding advantages over AM and K-means. The 12 typical days selected by K-MIMLP retained a large heat load, so the total heating capacity of the configuration results was close to the reference case. (3) Energy storage equipment: The energy storage potential of AM and K-means was underestimated due to the "smoothing effect". The typical daily heat loads of AM and K-means were smoother than those of K-MIMLP, especially AM. However, in order to retain extreme scenarios, K-MIMLP overestimated the energy storage potential to a certain extent, resulting in a larger capacity of energy storage equipment. However, for energy system design, redundant design is better than insufficient design. And the energy storage capacity deviation of K-MIMLP is still the smallest. In summary, compared with the reference case, the accuracy of the planning schemes is as follows: K-MIMLP > K-means > AM.

[0056] The annual costs and annual energy consumption are shown in Table 3. It can be seen that compared with the reference case, the annual total costs of typical days are lower than those of the reference case. Obviously, the optimization target value of K-MLMLP has the highest accuracy.

[0057]

[0058] Table 3

[0059] S7. Use peak index, cumulative index and morphological index to evaluate its accuracy. The evaluation indexes established are as follows:

[0060] S71 daily peak deviation

[0061]

[0062] It represents the time series value of attribute a (eg: cold, heat, electric load, solar radiation, temperature, etc.) on the i-th day, the h-th hour, after the original time series is replaced by the typical daily time series. Represents the original time series value. smax[x, z(x, y)] represents the maximum value of z(x, y) under the x index. represents the daily peak value of a typical daily time series, represents the daily peak value of the original time series. Therefore, Δ DP Describes the degree to which the typical daily time series retains the daily peak value of the original time series.

[0063] S72 Days Worst Condition Deviation

[0064]

[0065] If the attribute is not unique, it is necessary to find the most unfavorable operating point. (A) The most unfavorable operating point on one side of the load: The time series contains multiple loads. The first term in the numerator of formula (15) represents the total daily peak load of a typical daily time series, the second term represents the total daily peak load of the original time series. Therefore, Describes the most unfavorable operating condition deviation on one side of the load between the typical day and the original time series. (B) The most unfavorable operating point on both sides of the source and load: The time series contains not only multiple loads but also renewable energy. The first term in the numerator of formula (16) The second term represents the daily peak value of the net load of a typical daily time series. represents the daily peak value of the net load of the original time series. Therefore, The worst operating deviations on both the source and load sides between the typical day and the original time series are described.

[0066] S73 Annual cumulative value deviation

[0067]

[0068] represents the annual cumulative value of a typical daily time series, Represents the annual cumulative value of the original time series. Δ AC Describes the deviation of the annual cumulative value of the time series between the typical day and the original data.

[0069] S74 daily pattern deviation

[0070]

[0071] represents the daily average of the typical daily time series and the original time series. Therefore, Δ DP It describes the deviation between the fluctuations of the typical daily time series and the original time series based on their respective means, reflecting the degree to which the typical day retains the daily change trend of the original time series.

[0072] S75 daily maximum rate of change deviation

[0073]

[0074] In formula (19), the first term in the numerator represents the maximum deviation of the typical daily time series in adjacent periods, and the second term represents the maximum deviation of the original time series in adjacent periods. DMCR It reflects the degree to which a typical day retains the maximum daily change rate of the original time series.

[0075] The evaluation index results obtained by different typical day selection methods are as follows: Figure 6 , the most unfavorable operating condition deviation on both sides of the source and load is the most critical. The most unfavorable operating condition deviations of AM, K-means, and K-MIMLP are 28.1%, 22.4%, and 10.6%, respectively. This shows that K-MIMLP improves the accuracy of typical days by setting the most unfavorable operating condition constraints. In general, the accuracy of the three typical day evaluation indicators from high to low is: K-MIMLP>K-means>AM.

[0076] The present invention is not limited to the embodiments described above. The above description of the specific embodiments is intended to describe and illustrate the technical solution of the present invention. The above specific embodiments are merely illustrative and not restrictive. Without departing from the scope of the present invention and the scope of protection of the claims, a person of ordinary skill in the art can also make many forms of specific changes under the guidance of the present invention, which all fall within the scope of protection of the present invention.

Claims

1. A typical day design boundary extraction method for an integrated energy system based on optimization, characterized in that: The following steps are involved: S1. Collect 8760 hours of original time series, including cooling load, heating load, electric load, and renewable energy output; S2, the data were standardized using the z-score standardization method; S3, eliminate extreme days, and each day when an attribute reaches the annual peak is considered an extreme day; S4. Set the typical days, target weight, and constraint parameters in sequence; S5, after reaching convergence, add the removed extreme days again; S6. Output typical day weight; S7. Use peak index, cumulative index, and pattern index to evaluate the accuracy of typical days; The optimization goal is to minimize the sum of the dissimilarity between the Euclidean distance and the Pearson correlation coefficient, and to impose constraints on extreme values, cumulative values, and morphology. The expression of the objective function is as follows: bin i,j is a binary variable. If the jth day is replaced by the ith day, bin i,j =1, otherwise bin i,j =0;d i，j represents the Euclidean distance between the i-th day and the j-th day; r i,j represents the Pearson correlation coefficient between day i and day j, 1-r i，j It represents the difference between the two, cov represents the covariance, and σ represents the standard deviation; V eu 、V pc is the corresponding target weight; δ is the dimensionless coefficient; The constraint expressions of this method are as follows: Formula (4) forces each day j to be replaced by only one typical day i; Formula (5) is used to calculate the weight W of the typical day i ; In formulas (6) to (7), is also a binary variable. If the i-th day is selected as a typical day, otherwise K is the number of typical days, which can be set by the user. Formulas (8) to (13) constrain the deviation between the extracted typical days and the original time series in terms of extreme values, cumulative values, and forms.

2. The typical day design boundary extraction method of the integrated energy system based on optimization according to claim 1 is characterized in that: The step S5 reaching convergence means that the objective function reaches a minimum value.

Images