FAST GENERATION ADJUSTMENT ALGORITHM FOR ENERGY MANAGEMENT SYSTEM

Abstract

A method is proposed for real-time economic dispatch in power system operation, especially in systems that include renewable power sources that may cause heavy deviations from a generation schedule due to their uncertain outputs. According to the method an optimal generation schedule is obtained based on forecasted load data. This schedule has to be adjusted if the actual load and renewable energy source outputs unexpectedly deviate from the forecasted value. An algorithm employed in the method is capable of dictating generation adjustments which minimize total generation costs. The algorithm is a modification of the base point and participation factor method. It differs from existing methods in that a precise model of transmission losses is adopted in the algorithm to achieve higher accuracy in optimization. The proposed method also has significant advantage in execution speed so that it is suitable for real-time operation.

Claims

1. In an electrical power system having buses, power generators and loads, a method of minimizing total generation cost in real-time economic dispatch within the power system, comprising the steps of: calculating an original power generation schedule for the power generators in the power system based on a forecast of the loads and the fuel cost of operating each generator; in real-time operation, in response to a deviation of load from the schedule, calculating by a computer in the energy management system the optimal generation adjustments to the original power generation schedule according to the algorithm $\hspace{1em}\begin{array}{c}\mathrm{\Delta \lambda }=\frac{\partial {\lambda }_1}{\partial P_{G.Math..Math.1}}\times \Delta .Math..Math.P_{G.Math..Math.1}+\frac{\partial {\lambda }_1}{\partial P_{G.Math..Math.2}}\times \Delta .Math..Math.P_{G.Math..Math.2}+\mathrm{.Math.}.Math..Math.\frac{\partial {\lambda }_1}{\partial P_{\mathrm{GNG}}}\times \Delta .Math..Math.P_{\mathrm{GNG}}\\ \mathrm{\Delta \lambda }=\frac{\partial {\lambda }_2}{\partial P_{G.Math..Math.1}}\times \Delta .Math..Math.P_{G.Math..Math.2}+\frac{\partial {\lambda }_2}{\partial P_{G.Math..Math.2}}\times \Delta .Math..Math.P_{G.Math..Math.2}+\mathrm{.Math.}.Math..Math.\frac{\partial {\lambda }_2}{\partial P_{\mathrm{GNG}}}\times \Delta .Math..Math.P_{\mathrm{GNG}}\\ \mathrm{.Math.}\\ \mathrm{\Delta \lambda }=\frac{\partial {\lambda }_{\mathrm{NG}}}{\partial P_{G.Math..Math.1}}\times \Delta .Math..Math.P_{G.Math..Math.1}+\frac{\partial {\lambda }_{\mathrm{NG}}}{\partial P_{G.Math..Math.2}}\times \Delta .Math..Math.P_{G.Math..Math.2}+\mathrm{.Math.}.Math..Math.\frac{\partial {\lambda }_{\mathrm{NG}}}{\partial P_{\mathrm{GNG}}}\times \Delta .Math..Math.P_{\mathrm{GNG}}\\ \Delta .Math..Math.P_{G.Math..Math.1}^{\prime }+\Delta .Math..Math.P_{G.Math..Math.2}^{\prime }+\mathrm{.Math.}.Math..Math.\Delta .Math..Math.P_{\mathrm{GNG}}^{\prime }=1\end{array}$ where Δλ is the change of equal incremental cost before and after the generation adjustment, ΔP.sub.Gi is the output adjustment for generator i, NG is the number of generators in the system; and changing the operation of the power generators through the energy management system in accordance with the calculated generation adjustments to the schedule.

2. The method of claim 1 wherein the load deviations comprise real-power injection changes at any bus of the system.

3. The method of claim 1 wherein the power system includes renewable power sources and fluctuation in the renewable power sources are regarded in the algorithm as negative load deviations.

4. The method of claim 1 wherein the equal incremental cost is the basic principle for the generation adjustment, equal incremental cost holds for every generator in the original generation schedule, and a different equal incremental cost should still hold for every generator after the adjustment; thus the change of the two equal incremental costs (Δλ) should be equal for all generators; and since Δλ can be expressed by the adjustments of generations (ΔP.sub.Gi) as shown in claim 1, the equation set can be used to solve ΔP.sub.Gi.

5. The method of claim 1 wherein the expression of equal incremental cost takes transmission losses into consideration, and the expression is shown as follow: ${\lambda }_i^0={\lambda }^0=\frac{\frac{{\mathrm{dF}}_i}{{\mathrm{dP}}_{\mathrm{Gi}}}}{\left(1-\frac{\partial P_L}{\partial P_{\mathrm{Gi}}}\right)}$ where λ.sup.0 is the original equal incremental cost, F.sub.i is the cost function of generator i, P.sub.Gi is the original generation scheduled for generator i, and P.sub.L is the transmission losses.

6. The method of claim 1 wherein the change of equal incremental costs (Δλ) is formulated as the following expression: $\mathrm{\Delta \lambda }=\frac{\partial {\lambda }_i^0}{\partial P_{G.Math..Math.1}}\times \Delta .Math..Math.P_{G.Math..Math.1}+\frac{\partial {\lambda }_i^0}{\partial P_{G.Math..Math.2}}\times \Delta .Math..Math.P_{G.Math..Math.2}+\mathrm{.Math.}+\frac{\partial {\lambda }_i^0}{\partial P_{\mathrm{GNG}}}\times \Delta .Math..Math.P_{\mathrm{GNG}}$ the index i represents generator i and the above equation holds for i=1 to NG; and by the expression of λ.sub.i.sup.0 in claim 5, ∂λ.sub.i.sup.0/∂P.sub.Gj can be calculated by: $\frac{\partial {\lambda }_i^0}{\partial P_{\mathrm{Gj}}}=\{\begin{array}{cc}\frac{\frac{\partial F_i}{\partial P_{\mathrm{Gi}}}}{{\left(1-\frac{\partial P_L}{\partial P_{\mathrm{Gi}}}\right)}^2}\times \frac{{\partial }^2.Math.P_L}{\partial P_{\mathrm{Gi}}.Math.\partial P_{\mathrm{Gj}}},& i\ne j\\ \frac{\frac{{\partial }^2.Math.F_i}{\partial P_{\mathrm{Gi}}^2}}{\left(1-\frac{\partial P_L}{\partial P_{\mathrm{Gi}}}\right)}+\frac{\frac{\partial F_i}{\partial P_{\mathrm{Gi}}}}{{\left(1-\frac{\partial P_L}{\partial P_{\mathrm{Gi}}}\right)}^2}\times \frac{{\partial }^2.Math.P_L}{\partial P_{\mathrm{Gi}}^2},& i=j\end{array}$ this is put back to the expression of Δλ to formulate the equation set in claim 1.

7. The method of claim 6 wherein second order approximation of transmission losses are adopted, and both first order and second order derivatives of transmission losses to generator′ outputs are calculated using a sensitivity method; and wherein a unit increment is set to each generator to calculate these derivatives, and slack of the increment of generation is distributed slack.

8. The method of claim 7 wherein the value of the distributed slack is a normalized vector of load deviation at each bus of the power system.

9. The method of claim 8 wherein the choice of the slack reflects the actual power flow change after the generation increment.

10. The method of claim 7 wherein the derivatives of transmission losses are calculated by the expressions: $.Math.\frac{\partial P_L}{\partial P_{\mathrm{Gi}}}=\underset{l=1}{\overset{\mathrm{NL}}{.Math.}}.Math..Math.\frac{\partial P_L}{\partial {\theta }_{\mathrm{mn}}}\times \frac{\partial {\theta }_{\mathrm{mn}}}{\partial P_{\mathrm{Gi}}}$ $.Math.\mathrm{and}$ $\frac{{\partial }^2.Math.P_L}{\partial P_{\mathrm{Gi}}.Math.\partial P_{\mathrm{Gj}}}=\underset{l=1}{\overset{\mathrm{NL}}{.Math.}}.Math..Math.\left[\frac{\partial \left(\underset{l=1}{\overset{\mathrm{NL}}{.Math.}}.Math..Math.\frac{\partial P_L}{\partial {\theta }_{\mathrm{mn}}}\times \frac{\partial {\theta }_{\mathrm{mn}}}{\partial P_{\mathrm{Gi}}}\right)}{\partial {\theta }_{\mathrm{mn}}}\times \frac{\partial {\theta }_{\mathrm{mn}}}{\partial P_{\mathrm{Gj}}}\right]=\underset{l=1}{\overset{\mathrm{NL}}{.Math.}}.Math..Math.\left(\frac{{\partial }^2.Math.P_L}{\partial {\theta }_{\mathrm{mn}}^2}\times \frac{\partial {\theta }_{\mathrm{mn}}}{\partial P_{\mathrm{Gi}}}\times \frac{\partial {\theta }_{\mathrm{mn}}}{\partial P_{\mathrm{Gj}}}\right)$ where NL is the number of transmission lines in the system and θ.sub.mn is the voltage angle difference at the two ends of the transmission line.

11. The method of claim 1, wherein generation adjustments are only calculated for generators not operating at their upper or lower limits.

12. The method of claim 1, wherein any generator whose output exceeds its upper or lower limit is caused to operate at its limit and the power exceeding the limit is compensated by other generators which have capacity, in proportion to the value of their generation adjustment results.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The foregoing and other objects and advantages of the present invention will become more readily apparent when considered in connection with the following detailed description and appended drawings in which like designations denote like elements in the various views, and wherein;

[0016] FIG. 1 is a diagram of the IEEE 30 Bus Test Case; and

[0017] FIG. 2 is a flow chart of the method of the present invention.

DETAILED DESCRIPTION OF AN ILLUSTRATIVE EMBODIMENT

[0018] The disclosed invention is a method for real-time economic dispatch that relies upon an algorithm that calculates the optimal generation adjustments when loads deviate from the optimal generation schedule.

[0019] Consider a power system with NB buses and NG generators. Assume that an optimal original generation schedule or base point has been given and the deviation of load from the base point is denoted by ΔP.sub.D, which is an NB×1 vector corresponding to each bus in the system. The deviation of renewable energy sources can be regarded as negative load deviation and put into the ΔP.sub.D vector. The optimal generation adjustment ΔP.sub.G, which is a NG×1 vector for each generator, is to be obtained.

[0020] By the classical economic dispatch principle, the incremental cost of all generators in the system will be equal at the base point. The equal incremental cost is expressed as λ.sup.0, which is also the Lagrange multiplier of the Lagrange function. The symbol λ.sup.0 involving the transmission losses can be expressed as follows:

[00001]$\begin{array}{cc}{\lambda }_i^0={\lambda }^0=\frac{\frac{{\mathrm{dF}}_i}{{\mathrm{dP}}_{\mathrm{Gi}}}}{\left(1-\frac{\partial P_L}{\partial P_{\mathrm{Gi}}}\right)}& \left(1\right)\end{array}$

[0021] After the load deviation, assume that the equal incremental cost moves from λ.sup.0 to λ.sup.1=λ.sup.0+Δλ at the new optimal point when the output vector of the generators changes from P.sub.G to P.sub.G+ΔP.sub.G. Expressing λ as a function of P.sub.G, by a Taylor expansion yields:

[00002]$\begin{array}{cc}{\lambda }^1={\lambda }^0+\mathrm{\Delta \lambda }={\lambda }^0+\frac{\partial {\lambda }_i^0}{\partial P_{G.Math..Math.1}}\times \Delta .Math..Math.P_{G.Math..Math.1}+\frac{\partial {\lambda }_i^0}{\partial P_{G.Math..Math.2}}\times \Delta .Math..Math.P_{G.Math..Math.2}+\mathrm{.Math.}+\frac{\partial {\lambda }_i^0}{\partial P_{\mathrm{GNG}}}\times \Delta .Math..Math.P_{\mathrm{GNG}}& \left(2\right)\end{array}$

[0022] Since the new λ.sup.1 is equal for all generators, Δλ should also be equal for all generators. Thus equation (3) holds for i=1 to NG:

[00003]$\begin{array}{cc}\mathrm{\Delta \lambda }=\frac{\partial {\lambda }_i^0}{\partial P_{G.Math..Math.1}}\times \Delta .Math..Math.P_{G.Math..Math.1}+\frac{\partial {\lambda }_i^0}{\partial P_{G.Math..Math.2}}\times \Delta .Math..Math.P_{G.Math..Math.2}+\mathrm{.Math.}+\frac{\partial {\lambda }_i^0}{\partial P_{\mathrm{GNG}}}\times \Delta .Math..Math.P_{\mathrm{GNG}}& \left(3\right)\end{array}$

[0023] The precise expression of ∂λ.sub.i/∂P.sub.Gj after involving transmission losses is:

[00004]$\begin{array}{cc}\frac{\partial {\lambda }_i^0}{\partial P_{\mathrm{Gj}}}=\{\begin{array}{cc}\frac{\frac{\partial F_i}{\partial P_{\mathrm{Gi}}}}{{\left(1-\frac{\partial P_L}{\partial P_{\mathrm{Gi}}}\right)}^2}\times \frac{{\partial }^2.Math.P_L}{\partial P_{\mathrm{Gi}}.Math.\partial P_{\mathrm{Gj}}},& i\ne j\\ \frac{\frac{{\partial }^2.Math.F_i}{\partial P_{\mathrm{Gi}}^2}}{\left(1-\frac{\partial P_L}{\partial P_{\mathrm{Gi}}}\right)}+\frac{\frac{\partial F_i}{\partial P_{\mathrm{Gi}}}}{{\left(1-\frac{\partial P_L}{\partial P_{\mathrm{Gi}}}\right)}^2}\times \frac{{\partial }^2.Math.P_L}{\partial P_{\mathrm{Gi}}^2},& i=j\end{array}& \left(4\right)\end{array}$

[0024] The original prior art participation factors method ignores the effects of transmission losses, namely P.sub.L, ∂P.sub.L/∂P.sub.Gi and ∂.sup.2P.sub.L(∂P.sub.Gi∂P.sub.Gj) are considered to be zero. However in the proposed algorithm, a precise model of transmission losses is used to calculate ∂.sub.PL/∂P.sub.Gi and ∂.sup.2P.sub.L/(∂P.sub.Gi∂P.sub.Gj). The model has two major characteristics: first it adopts a second order approximation, namely the second order derivatives of losses in generation are calculated, considering the non-linear expression of transmission losses. Second, the slack bus is not the commonly used single reference bus in the prior art transmission losses modeling, but is a particular distributed slack related to the load deviation ΔP.sub.D vector. These changes achieve a much more accurate modeling of transmission losses.

[0025] Assume that the cost functions F for generators are quadratic functions, which are most commonly adopted. The derivatives to F are easy to calculate. The questions are how to calculate the first and second order derivatives of transmission losses to the outputs of the generators.

[0026] The disclosed algorithm adopts a second order approximation of the transmission losses and distributed slack in the modeling. The first order derivatives of losses are calculated using (5):

[00005]$\begin{array}{cc}\frac{\partial P_L}{\partial P_{\mathrm{Gi}}}=\underset{l=1}{\overset{\mathrm{NL}}{.Math.}}.Math..Math.\frac{\partial P_L}{\partial {\theta }_{\mathrm{mn}}}\times \frac{\partial {\theta }_{\mathrm{mn}}}{\partial P_{\mathrm{Gi}}}& \left(5\right)\end{array}$

[0027] The term ∂P.sub.L/∂θ.sub.mn can be calculated by the precise expression of transmission losses:

P.sub.L=Σ.sub.i=1.sup.NLG.sub.mn×(V.sub.m.sup.2+V.sub.n.sup.2−2V.sub.mV.sub.n×cos θ.sub.mn)  (6)

[0028] In equation (6), NL is the total number of transmission lines in the system, m and n are the bus indexes at the two ends of line i, θ.sub.mn is the voltage angle difference at the two ends of line i, V.sub.m and V.sub.n are the voltage magnitudes at the two ends of line i, and G.sub.mn is the conductance of line i.

[0029] The term ∂θ.sub.mn/∂P.sub.Gi can be calculated using sensitivity method with DC power flow model:

Δθ=X×ΔP  (7)

[0030] The ΔP vector is set to a unit increment for generator i plus the distributed slack which is equal to the normalized vector of load deviation:

[00006]$\begin{array}{cc}\Delta .Math..Math.P\left(j\right)=\{\begin{array}{cc}-1,& j=i\\ {\alpha }_k,& j=k,\mathrm{any}.Math..Math.\mathrm{buses}.Math..Math.\mathrm{with}.Math..Math.\mathrm{load}.Math..Math.\mathrm{deviation}\\ 0,& \mathrm{otherwise}\end{array}& \left(8\right)\end{array}$

And:

[0031] [00007]$\begin{array}{cc}{\alpha }_k=\frac{\Delta .Math..Math.P_D\left(k\right)}{\mathrm{\Sigma \Delta }.Math..Math.P_d}& \left(9\right)\end{array}$

[0032] Notice that this should be calculated for each generator i=1 to NG. Since ΔP.sub.Gi=1, ∂θ.sub.mn/∂P.sub.Gi is equivalent to (Δθ.sub.n−Δθ.sub.m) using the result of Δθ.

[0033] The second order derivatives are calculated using (10). ∂θ.sub.mn/∂P.sub.Gi is considered to be constant for a specific case, which is reasonable because the relationship between voltage angle and active power injection is approximately linear.

[00008]$\begin{array}{cc}\frac{{\partial }^2.Math.P_L}{\partial P_{\mathrm{Gi}}.Math.\partial P_{\mathrm{Gj}}}=\underset{l=1}{\overset{\mathrm{NL}}{.Math.}}.Math..Math.\left[\frac{\partial \left(\underset{l=1}{\overset{\mathrm{NL}}{.Math.}}.Math..Math.\frac{\partial P_L}{\partial {\theta }_{\mathrm{mn}}}\times \frac{\partial {\theta }_{\mathrm{mn}}}{\partial P_{\mathrm{Gi}}}\right)}{\partial {\theta }_{\mathrm{mn}}}\times \frac{\partial {\theta }_{\mathrm{mn}}}{\partial P_{\mathrm{Gj}}}\right]=\underset{l=1}{\overset{\mathrm{NL}}{.Math.}}.Math..Math.\left(\frac{{\partial }^2.Math.P_L}{\partial {\theta }_{\mathrm{mn}}^2}\times \frac{\partial {\theta }_{\mathrm{mn}}}{\partial P_{\mathrm{Gi}}}\times \frac{\partial {\theta }_{\mathrm{mn}}}{\partial P_{\mathrm{Gj}}}\right)& \left(10\right)\end{array}$

[0034] Put the results of the derivatives into (4), then put (4) into (3). The equations become:

[00009]$.Math.\left(11\right)$$\begin{array}{c}\mathrm{\Delta \lambda }=\frac{\partial {\lambda }_1}{\partial P_{G.Math..Math.1}}\times \Delta .Math..Math.P_{G.Math..Math.1}+\frac{\partial {\lambda }_1}{\partial P_{G.Math..Math.2}}\times \Delta .Math..Math.P_{G.Math..Math.2}+\mathrm{.Math.}.Math..Math.\frac{\partial {\lambda }_1}{\partial P_{\mathrm{GNG}}}\times \Delta .Math..Math.P_{\mathrm{GNG}}\\ \mathrm{\Delta \lambda }=\frac{\partial {\lambda }_2}{\partial P_{G.Math..Math.1}}\times \Delta .Math..Math.P_{G.Math..Math.2}+\frac{\partial {\lambda }_2}{\partial P_{G.Math..Math.2}}\times \Delta .Math..Math.P_{G.Math..Math.2}+\mathrm{.Math.}.Math..Math.\frac{\partial {\lambda }_2}{\partial P_{\mathrm{GNG}}}\times \Delta .Math..Math.P_{\mathrm{GNG}}\\ \mathrm{.Math.}\\ \mathrm{\Delta \lambda }=\frac{\partial {\lambda }_{\mathrm{NG}}}{\partial P_{G.Math..Math.1}}\times \Delta .Math..Math.P_{G.Math..Math.1}+\frac{\partial {\lambda }_{\mathrm{NG}}}{\partial P_{G.Math..Math.2}}\times \Delta .Math..Math.P_{G.Math..Math.2}+\mathrm{.Math.}.Math..Math.\frac{\partial {\lambda }_{\mathrm{NG}}}{\partial P_{\mathrm{GNG}}}\times \Delta .Math..Math.P_{\mathrm{GNG}}\\ \Delta .Math..Math.P_{G.Math..Math.1}^{\prime }+\Delta .Math..Math.P_{G.Math..Math.2}^{\prime }+\mathrm{.Math.}.Math..Math.\Delta .Math..Math.P_{\mathrm{GNG}}^{\prime }=1\end{array}$

[0035] The solution to the above equations is the normalized generation adjustment factors ΔP.sub.G′, since the summation of the vector is set to one. This vector will be used for further adjustments due to system constraints and so on. If the system constraints are not considered, the generation adjustment result can be calculated by multiplying the generation adjustment factors with the total load deviation plus the change of transmission losses. The change of loss can be estimated as:

[00010]$\begin{array}{cc}\Delta .Math..Math.P_L=\mathrm{\Sigma \Delta }.Math..Math.P_D\times \Delta .Math..Math.P_G^{\prime }\times \frac{\partial P_L}{\partial P_G}& \left(12\right)\end{array}$

[0036] The generation adjustment result is obtained by:

[00011]$\begin{array}{cc}\Delta .Math..Math.P_G=\Delta .Math..Math.P_G^{\prime }\times \left(\mathrm{\Sigma \Delta }.Math..Math.P_D+\Delta .Math..Math.P_L\right)=\Delta .Math..Math.P_G^{\prime }\times \left(\mathrm{\Sigma \Delta }.Math..Math.P_D+\mathrm{\Sigma \Delta }.Math..Math.P_D\times \Delta .Math..Math.P_G^{\prime }\times \frac{\partial P_L}{\partial P_G}\right)& \left(13\right)\end{array}$

[0037] Secondary adjustments are needed to ensure that the power system constraints are not violated after the adjustment. These adjustments include:

[0038] (1) Only generators which are not operating at their upper or lower limits are involved in the calculation to ensure equal incremental cost. This is feasible for small adjustments around the local point;

[0039] (2) Adjust the outputs of the generators if any of them exceed the generators' upper or lower limits. The power which exceeds the limit should be compensated by other generators which have capacity, in proportion to their generation adjustment factors;

[0040] (3) Adjust the outputs of the generators if any transmission line is overloaded.

[0041] The generation adjustment methodology of the present invention generates the optimal adjustments to generation schedules in real-time economic dispatch, considering the fast fluctuations caused by the loads and renewable power sources. Whenever there is any deviation from the generation schedule, the present invention is able to calculate the adjusted outputs of the generators to obtain minimal total generation cost.

[0042] The optimization is effective and reliable, and the execution speed is very fast so as to fulfill the need of real-time operation.

[0043] A test of the present invention demonstrates that it can obtain optimal generation adjustment results with about 3 to 5 times smaller average error compared to the participation factors method. Further, its execution speed is more than 20 times faster than the optimal power flow (OPF) method, which is commonly used for the economic dispatch, in practical large power systems.

[0044] In order to test the present invention, simulations were executed in the IEEE 30-bus system and the IEEE 118-bus system. The base point result was calculated first and then the generation adjustments were calculated after a load deviation. The results compared include the total generation cost after the adjustment, and the errors of vector ΔP.sub.G compared to the Alternating Current Optimal Power Flow (AC-OPF) result, which is considered to be the true optimal result.

[0045] The IEEE 30 Bus System is shown in FIG. 1, which consists of 6 generators 1, 2, 13, 22, 23 and 27. In the IEEE 30 bus system at the base point, the optimal generation cost calculated by the AC-OPF method was 8906.15/hr. The method of the present invention, the OPF and the original participation method were used to calculate the optimal generation output when all the loads were increased by 10%. The results are shown in Table 1.

TABLE-US-00001 TABLE I GENERATION ADJUSTMENT RESULTS FOR IEEE 30-BUS SYSTEM Generation Average Cost Error Max Error Method ($/hr) (MW) (MW) AC-OPF 10053 — — Present Invention 10053 0.06 0.16 Participation factors 10055 2.17 6.58

[0046] As can be seen from Table 1, the average error for the present invention is very low compared to the conventional participation factors.

[0047] The IEEE 118 bus system (not shown) contains 54 generators, 186 lines and 99 load buses. This system has a lot of voltage control devices, and is quite robust, converging in 5 or so iterations with a fast-decoupled power flow. With the IEEE 118-bus system, the total generation cost at base point is 129660.69/hr. The load deviation with a 5% load increase is shown in Table II.

TABLE-US-00002 TABLE II GENERATION ADJUSTMENT RESULTS (5% LOAD INCREASE) Generation Average Cost Error Max Error Method ($/hr) (MW) (MW) AC-OPF 138090 — — Present Invention 138092 0.23 3.24 Participation factors 138100 1.25 6.11
It can be seen from Table II that the generation cost obtained by the present invention is lower than the participation factor method, and very close to the OPF result. The errors of the present invention are also much lower than the participation factor method.

[0048] Table III shows the load deviation for a 10% increase in load.

TABLE-US-00003 TABLE III GENERATION ADJUSTMENT RESULTS (10% LOAD INCREASE) Generation Average Cost Error Max Error Method ($/hr) (MW) (MW) AC-OPF 146572 — — Present Invention 146578 0.77 5.54 Participation factors 146610 2.90 11.80
Again, the generation cost and errors by the present invention are much lower than the participation factor method.

[0049] A comparison of the execution speed or time for the present invention to that of the AC-OPP method is shown in Table IV. This comparison of execution time is for the IEEE 118-bus case.

TABLE-US-00004 TABLE IV Excution Time of AC-OPF and the Proposed Methond Test system CPU time (sec) 118-bus system Present Invention 0.015 AC-OPF 0.32
The response time of the current invention is 20 times faster than the OPF method.

[0050] The current invention uses a precise second-order model for the transmission losses in the power system, and a distributed slack to access the effect of transmission losses to the generation adjustment factors such that the generation adjustment algorithm is accurate and fast. Small increases in efficiency of dispatch are measured in billions of dollars per year. Since the usual cost of purchasing and installing new software for an existing ISO market is less than 10 million dollars, the potential benefit/cost ratios of better software are in the range of 10 to 1000.

[0051] The key idea of the present invention is that the transmission losses will affect the participation factors themselves, and a precise model of transmission losses is used to calculate the new participation factors, namely the “generation adjustment” in the method of the present invention. Further, the proposed method is used to calculate generation adjustments when the load is deviated from the original economic dispatch plan. Different load deviation will lead to a particular distributed slack in the transmission losses modeling.

[0052] FIG. 2 is a flow diagram showing the whole process or method of the present invention. This process is executed at any time point when load deviation from the generation schedule occurs. The process assumes that an optimal generation schedule or base point exists. If is does not exist, the base point is calculated in the conventional manner based on a forecast of the expected load and the goal of achieving equal incremental costs for the system generators. The schedule is executed by computer control of the generating capacity through a standard Energy Management System (EMS).

[0053] The first step 201 of the method is to read the base point data and date describing the power distribution system that is being controlled. This data is input to the EMS computer system, which will also execute the algorithm according to the present invention. Next, load deviations ΔP.sub.D are sensed and in step 203 the load deviation information is entered into the system computer. The computer system in step 205 than calculates the first order derivative ∂P.sub.L/∂P.sub.Gi according to equation (5) for each generator 1 to NG. Similarly, the computer in step 207 calculates the second order derivative ∂.sup.2P.sub.L/∂P.sub.Gi ∂P.sub.Gj according to equation (10) for generators i=1 to NG and j=1 to NG.

[0054] In step 209 the computer calculates ∂λ.sub.i/∂P.sub.Gj using equation (4) for i=1 to NG and j=1 to NG. Then in step 211 equations (11) are solved for the generation adjustment factors. Each of the generation adjustment factors is multiplied by the total load deviation to obtain the generation adjustment results in step 213. Finally, in step 215 the results are modified to satisfy the constraints, if necessary, in order to obtain the final generation adjustment result.

[0055] The present invention can handle most major power system constraints. The algorithm is to be used by being programmed and integrated to existing EMS systems.

[0056] While the invention has been particularly shown and described with reference to a preferred embodiment thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention.