I am wondering if my implementation of voltage sensitivity coefficients for unbalanced three-phase-four-wire radial distribution feeders is correct.

We got \$\bar{V}\$ for phase(s) of all the buses and \$\bar{Y}\$, the admittance matrix of the feeder. \$\mathcal{S}\$ is the set for phases of the slack bus, and \$\mathcal{N}\$ is the set for phase(s) of all the buses. For real and reactive power injections \$P\$, \$Q\$, all wye-connected,the following relationship can be obtained using first order Taylor expansion: $$ d |\bar{V}_{i}| = \sum_{j \in \mathcal{N}} \frac{\partial |\bar{V}_{i}|}{\partial P_j} \cdot d P_j + \sum_{j \in \mathcal{N}} \frac{\partial |\bar{V}_i|}{\partial Q_j} \cdot d Q_j \quad \forall i \in \mathcal{N} $$ where we should get \$\partial \bar{V}_{i} / \partial P_{j} \$ first, and then use: $$ \frac{\partial\left|\bar{V}_{i}\right|}{\partial P_{l}}=\frac{1}{\left|\bar{V}_{i}\right|} \operatorname{Re}\left(\underline{V}_{i} \frac{\partial \bar{V}_{i}}{\partial P_{l}}\right) $$

The calculation for \$\partial \bar{V}_{i} / \partial Q_j\$ is similar.

In [christakou2013efficient], to calculate \$\partial V_{i} / \partial P_j\$, \$(\forall i, j \in \mathcal{N} \cap \mathcal{S}) \$, we need to solve the following set of linear equations: $$ \begin{aligned} \frac{\partial \bar{V}_{i}}{\partial P_{k}} &= 0 \quad \forall i \in \mathcal{S}, \forall k \in \mathcal{S} \cup \mathcal{N} \\ \frac{\partial \underline{V}_{i}}{\partial P_{k}} &= 0 \quad \forall i \in \mathcal{S}, \forall k \in \mathcal{S} \cup \mathcal{N} \\ \frac{\partial \bar{V}_{i}}{\partial P_{k}} &= 0 \quad \forall i \in \mathcal{N}, \forall k \in \mathcal{S} \\ \frac{\partial \underline{V}_{i}}{\partial P_{k}} &= 0 \quad \forall i \in \mathcal{N}, \forall k \in \mathcal{S} \\ \frac{\partial \underline{V}_{i}}{\partial P_{i}} \sum_{j \in \mathcal{S} \cup \mathcal{N}} \mathrm{Re}(\bar{Y}_{i j} \bar{V}_j) + \mathrm{Im}(\underline{V}_i) \sum_{j \in \mathcal{N}} \mathrm{Im}(\bar{Y}_{i j}) \frac{\partial \bar{V}_{j}}{\partial P_i} + \mathrm{Re}(\underline{V}_i) \sum_{j \in \mathcal{N}} \mathrm{Re}(\bar{Y}_{i j}) \frac{\partial \bar{V}_{j}}{\partial P_i} &= 1 \quad \forall i \in \mathcal{N} \\ \frac{\partial \underline{V}_{i}}{\partial P_{i}} \sum_{j \in \mathcal{S} \cup \mathcal{N}} \mathrm{Im}(\bar{Y}_{i j} \bar{V}_j) + \mathrm{Im}(\underline{V}_i) \sum_{j \in \mathcal{N}} \mathrm{Re}(\bar{Y}_{i j}) \frac{\partial \bar{V}_{j}}{\partial P_i} + \mathrm{Re}(\underline{V}_i) \sum_{j \in \mathcal{N}} \mathrm{Im}(\bar{Y}_{i j}) \frac{\partial \bar{V}_{j}}{\partial P_i} &= 0 \quad \forall i \in \mathcal{N} \\ \frac{\partial \underline{V}_{i}}{\partial P_k} \sum_{j \in \mathcal{S} \cup \mathcal{N}} \mathrm{Re}(\bar{Y}_{i j} \bar{V}_j) + \mathrm{Im}(\underline{V}_i) \sum_{j \in \mathcal{N}} \mathrm{Im}(\bar{Y}_{i j}) \frac{\partial \bar{V}_{j}}{\partial P_k} + \mathrm{Re}(\underline{V}_i) \sum_{j \in \mathcal{N}} \mathrm{Re}(\bar{Y}_{i j}) \frac{\partial \bar{V}_{j}}{\partial P_k} &= 0 \quad \forall i, k \in \mathcal{N}, k \neq i \\ \frac{\partial \underline{V}_{i}}{\partial P_k} \sum_{j \in \mathcal{S} \cup \mathcal{N}} \mathrm{Im}(\bar{Y}_{i j} \bar{V}_j) + \mathrm{Im}(\underline{V}_i) \sum_{j \in \mathcal{N}} \mathrm{Re}(\bar{Y}_{i j}) \frac{\partial \bar{V}_{j}}{\partial P_k} + \mathrm{Re}(\underline{V}_i) \sum_{j \in \mathcal{N}} \mathrm{Im}(\bar{Y}_{i j}) \frac{\partial \bar{V}_{j}}{\partial P_k} &= 0 \quad \forall i, k \in \mathcal{N}, k \neq i \end{aligned} $$ where \$\underline{V}\$ is the conjugate of \$\bar{V}\$. I am not sure about this equation set, because it is written in a simplified version in the paper. In [christakou2013efficient], there are only two equations: $$ \begin{align} \frac{\partial \bar{V}_{i}}{\partial P_{l}} &= 0, \quad \forall i \in \mathcal{S} \\ \frac{\partial \underline{V}_{i}}{\partial P_{l}} \sum_{j \in \mathcal{S} \cup \mathcal{N}} \bar{Y}_{i j} \bar{V}_{j}+\underline{V}_{i} \sum_{j \in \mathcal{N}} \bar{Y}_{i j} \frac{\partial \bar{V}_{j}}{\partial P_{l}} &= \mathbb{1}_{\{i=l\}} \end{align} $$

Besides, I am wondering if there are other methods. I read some method using impedance matrix, but I don't want delta-connected loads in [maharjan2020enhanced]. I know we cannot have inverted Jacobian matrix because Newton-Raphson algorithm should not be used. I don't want to use perturb-and-observe method, for example in [tamp2014sensitivity].

  • Christakou, K., LeBoudec, J. Y., Paolone, M., & Tomozei, D. C. (2013). Efficient computation of sensitivity coefficients of node voltages and line currents in unbalanced radial electrical distribution networks. IEEE Transactions on Smart Grid, 4(2), 741-750.
  • Maharjan, S., Khambadkone, A. M., & Peng, J. C. H. (2020). Enhanced Z-bus method for analytical computation of voltage sensitivities in distribution networks. IET Generation, Transmission & Distribution, 14(16), 3187-3197.
  • Tamp, F., & Ciufo, P. (2014). A sensitivity analysis toolkit for the simplification of MV distribution network voltage management. IEEE Transactions on Smart Grid, 5(2), 559-568.

1 Answer 1


I implemented another method from [zhou2008simplified]. Though the method is designed for balanced networks, I tried anyway.

The set of equations is: $$ \begin{aligned} \frac{\partial \bar{V}_{i}}{\partial P_{k}} &= \sum_{j \in \mathcal{N} / \{k\}} \Re \left[\frac{-\bar{Z}_{i j} \underline{S}_{j}}{\left(\underline{V}_{j}\right)^{2}} \right] \frac{\partial \underline{V}_{j}}{\partial P_{k}} + \Re \left[\frac{\bar{Z}_{i k} \underline{V}_{k}}{\left(\underline{V}_{k} \right)^{2}} \right] \quad \forall i, k \in \mathcal{N} \\ 0 &= \sum_{j \in \mathcal{N} / \{k\}} \Im \left[\frac{-\bar{Z}_{i j} \underline{S}_{j}}{\left(\underline{V}_{j}\right)^{2}} \right] \frac{\partial \underline{V}_{j}}{\partial P_{k}} + \Im \left[\frac{\bar{Z}_{i k} \underline{V}_{k}}{\left(\underline{V}_{k} \right)^{2}} \right] \quad \forall i, k \in \mathcal{N} \\ \end{aligned} $$ where impedance matrix, \$\bar{Z}\$, and complex power injections, \$\bar{S}\$, are used, which can be calculated using: $$ \begin{aligned} \bar{Z} &= \bar{Y}^{-1} \\ \bar{S} &= \bar{V} \odot \underline{I} \end{aligned} $$

But I couldn't get any solution so far.

I am implementing the method in [zad2018new], which I believe is based on fixed-point linearization (FPL). FPL is different from first-order Taylor expansion (FOT) used in [christakou2013efficient] and [zhou2008simplified], according to [bernstein2018load].

  • Zhou, Q., & Bialek, J. (2008, July). Simplified calculation of voltage and loss sensitivity factors in distribution networks. In Proc. 16th Power Syst. Comput. Conf.(PSCC2008).
  • Zad, B. B., Lobry, J., & Vallée, F. (2018). A New Voltage Sensitivity Analysis Method for Medium-Voltage Distribution Systems Incorporating Power Losses Impact. Electric Power Components and Systems, 46(14-15), 1540-1553.
  • Bernstein, A., Wang, C., Dall’Anese, E., Le Boudec, J. Y., & Zhao, C. (2018). Load flow in multiphase distribution networks: Existence, uniqueness, non-singularity and linear models. IEEE Transactions on Power Systems, 33(6), 5832-5843.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.